From andrew.nere at gmail.com Tue Jun 5 19:53:12 2012 From: andrew.nere at gmail.com (Andrew Nere) Date: Tue, 5 Jun 2012 12:53:12 -0500 Subject: [ilastik-user] Using/installing Ilastik Synapse Detection on Mac OS X 10.7 Lion Message-ID: Hello, I am attempting to use Ilastik Synapse Detection on Mac OS X 10.7 Lion. When I run the 'run-ilastik.sh' script, I get the following: dyld: Symbol not found: _objc_release Referenced from: /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation Expected in: /Users/nere/Desktop/ilastik-synapse-0.5.05-mac64/lib/libobjc.A.dylib in /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation ./run-ilastik.sh: line 3: 5249 Trace/BPT trap: 5 PYTHONPATH=.:$PYTHONPATH bin/python -B ilastik/ilastikMain.py I have been able to run the program with no problems on OS X 10.6 Snow Leopard with no problems. Anyone else run into these problems, or used Synapse Detection on Lion? Thanks Andy -------------- next part -------------- An HTML attachment was scrubbed... URL: From bergs at janelia.hhmi.org Tue Jun 5 20:17:09 2012 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Tue, 5 Jun 2012 18:17:09 +0000 Subject: [ilastik-user] Using/installing Ilastik Synapse Detection on Mac OS X 10.7 Lion In-Reply-To: References: Message-ID: An HTML attachment was scrubbed... URL: From bms2156 at columbia.edu Wed Aug 8 20:41:35 2012 From: bms2156 at columbia.edu (Benjamin Shababo) Date: Wed, 8 Aug 2012 13:41:35 -0500 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120730154022.16ovwol9w8gw4g8c@cubmail.cc.columbia.edu> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> <20120730154022.16ovwol9w8gw4g8c@cubmail.cc.columbia.edu> Message-ID: Is anyone available to follow up on this thread? It would be greatly appreciated, and also it would mean that our lab could use your software in our experiments! On Mon, Jul 30, 2012 at 2:40 PM, Ben Shababo wrote: > Anna- > > Also, it may be helpful to know what Linux platform/system was used to > develop the Linux version of Ilastik w/ Synapse Detection. > > Thanks! > > -Ben > > > Quoting anna.kreshuk at iwr.uni-**heidelberg.de > : > > Dear Mr. Shababo, >> >> Is this problem very urgent? I think I know where it comes from, but >> I'm only returning from vacation on Wednesday. Mayavi shouldn't be >> needed anymore, but the linux binary seems to correspond to an older >> synapse detection workflow version than the Mac binary. 0 synapses >> found is suspicious, but it could come from re-initialization of the cc >> volume. >> >> Sorry for the inconvenience >> Anna >> >> >> Quoting Ben Shababo : >> >> Update: >>> >>> We were able to solve the problem when filtering synapses... or at >>> least it doesn't throw an error now... we achieved this by calling >>> >>> cc = ScalarVolume(cc) >>> >>> which gives cc an axistags property. So far however, the algorithm >>> always returns 0 synapses. >>> >>> Also, we have another problem... when we go to visualize our selected >>> synapses (from the CC Result overlay, not CC Filtered), we get an error >>> that it cannot find the import for: >>> >>> Loading vtk ... >>> Traceback (most recent call last): >>> File >>> "/home/gt/Desktop/ilastik-0.5.**04-sd_fix/ilastik/gui/** >>> overlayWidget.py", >>> line 206, in onContext >>> from mayaviWidget import MayaviQWidget >>> File >>> "/home/gt/Desktop/ilastik-0.5.**04-sd_fix/ilastik/gui/** >>> mayaviWidget.py", >>> line 6, in >>> from enthought.mayavi.core.api import PipelineBase >>> ImportError: No module named mayavi.core.api >>> >>> I've tried putting different .egg folders into the site-packages, but >>> no luck. >>> >>> Thanks for any help. >>> ______________________________**_________________ >>> ilastik-user mailing list >>> ilastik-user at ilastik.org >>> http://mailman.ilastik.org/**cgi-bin/mailman/listinfo/**ilastik-user >>> >> > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Wed Aug 8 21:47:28 2012 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Wed, 08 Aug 2012 21:47:28 +0200 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> <20120730154022.16ovwol9w8gw4g8c@cubmail.cc.columbia.edu> Message-ID: <20120808214728.2iyish2p0ggcwo48@webmail.iwr.uni-heidelberg.de> Dear Ben, The axis tags problem can be solved by transforming the "cc" into a normal numpy array. If you call cc=numpy.asarray(cc) it won't be treated as a vigra array anymore and no axis tags will be required. I checked it today and there should be synapses found after applying this filter. However, you might want to re-check your size limits: for the small test dataset they should be something like 100 - 2000. The reason for this is that the small dataset is downsampled and the defaults correspond to the original resolution. Now for the Mayavi/VTK conflict, I'm planning to fix that tomorrow. I suppose that after messing so much with our code, you wouldn't mind getting a not-yet-official, but hopefully better working binary? Cheers, Anna Quoting Benjamin Shababo : > Is anyone available to follow up on this thread? It would be grfeatly > appreciated, and also it would mean that our lab could use your software in > our experiments! > > > On Mon, Jul 30, 2012 at 2:40 PM, Ben Shababo wrote: > >> Anna- >> >> Also, it may be helpful to know what Linux platform/system was used to >> develop the Linux version of Ilastik w/ Synapse Detection. >> >> Thanks! >> >> -Ben >> >> >> Quoting >> anna.kreshuk at iwr.uni-**heidelberg.de >> : >> >> Dear Mr. Shababo, >>> >>> Is this problem very urgent? I think I know where it comes from, but >>> I'm only returning from vacation on Wednesday. Mayavi shouldn't be >>> needed anymore, but the linux binary seems to correspond to an older >>> synapse detection workflow version than the Mac binary. 0 synapses >>> found is suspicious, but it could come from re-initialization of the cc >>> volume. >>> >>> Sorry for the inconvenience >>> Anna >>> >>> >>> Quoting Ben Shababo : >>> >>> Update: >>>> >>>> We were able to solve the problem when filtering synapses... or at >>>> least it doesn't throw an error now... we achieved this by calling >>>> >>>> cc = ScalarVolume(cc) >>>> >>>> which gives cc an axistags property. So far however, the algorithm >>>> always returns 0 synapses. >>>> >>>> Also, we have another problem... when we go to visualize our selected >>>> synapses (from the CC Result overlay, not CC Filtered), we get an error >>>> that it cannot find the import for: >>>> >>>> Loading vtk ... >>>> Traceback (most recent call last): >>>> File >>>> "/home/gt/Desktop/ilastik-0.5.**04-sd_fix/ilastik/gui/** >>>> overlayWidget.py", >>>> line 206, in onContext >>>> from mayaviWidget import MayaviQWidget >>>> File >>>> "/home/gt/Desktop/ilastik-0.5.**04-sd_fix/ilastik/gui/** >>>> mayaviWidget.py", >>>> line 6, in >>>> from enthought.mayavi.core.api import PipelineBase >>>> ImportError: No module named mayavi.core.api >>>> >>>> I've tried putting different .egg folders into the site-packages, but >>>> no luck. >>>> >>>> Thanks for any help. >>>> ______________________________**_________________ >>>> ilastik-user mailing list >>>> ilastik-user at ilastik.org >>>> http://mailman.ilastik.org/**cgi-bin/mailman/listinfo/**ilastik-user >>>> >>> >> >> > From bms2156 at columbia.edu Thu Aug 9 03:26:57 2012 From: bms2156 at columbia.edu (Benjamin Shababo) Date: Wed, 8 Aug 2012 20:26:57 -0500 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120808214728.2iyish2p0ggcwo48@webmail.iwr.uni-heidelberg.de> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> <20120730154022.16ovwol9w8gw4g8c@cubmail.cc.columbia.edu> <20120808214728.2iyish2p0ggcwo48@webmail.iwr.uni-heidelberg.de> Message-ID: Anna- Yes, that would be great, please send along any new versions of the code... we'd be happy to test! Also, thanks for the tip off on the size limit values, I'll try that and your suggestion of converting cc into a numpy array tomorrow when I'm back in the lab. -Ben On Wed, Aug 8, 2012 at 2:47 PM, wrote: > Dear Ben, > > The axis tags problem can be solved by transforming the "cc" into a normal > numpy array. If you call > cc=numpy.asarray(cc) > it won't be treated as a vigra array anymore and no axis tags will be > required. I checked it today and there should be synapses found after > applying this filter. However, you might want to re-check your size limits: > for the small test dataset they should be something like 100 - 2000. The > reason for this is that the small dataset is downsampled and the defaults > correspond to the original resolution. > > Now for the Mayavi/VTK conflict, I'm planning to fix that tomorrow. I > suppose that after messing so much with our code, you wouldn't mind getting > a not-yet-official, but hopefully better working binary? > > Cheers, > Anna > > Quoting Benjamin Shababo : > > Is anyone available to follow up on this thread? It would be grfeatly >> >> appreciated, and also it would mean that our lab could use your software >> in >> our experiments! >> >> >> On Mon, Jul 30, 2012 at 2:40 PM, Ben Shababo >> wrote: >> >> Anna- >>> >>> Also, it may be helpful to know what Linux platform/system was used to >>> develop the Linux version of Ilastik w/ Synapse Detection. >>> >>> Thanks! >>> >>> -Ben >>> >>> >>> Quoting anna.kreshuk at iwr.uni-**heidelb**erg.de < >>> anna.kreshuk at iwr.uni-**heidelberg.de >>> > >>> : >>> >>> Dear Mr. Shababo, >>> >>>> >>>> Is this problem very urgent? I think I know where it comes from, but >>>> I'm only returning from vacation on Wednesday. Mayavi shouldn't be >>>> needed anymore, but the linux binary seems to correspond to an older >>>> synapse detection workflow version than the Mac binary. 0 synapses >>>> found is suspicious, but it could come from re-initialization of the cc >>>> volume. >>>> >>>> Sorry for the inconvenience >>>> Anna >>>> >>>> >>>> Quoting Ben Shababo : >>>> >>>> Update: >>>> >>>>> >>>>> We were able to solve the problem when filtering synapses... or at >>>>> least it doesn't throw an error now... we achieved this by calling >>>>> >>>>> cc = ScalarVolume(cc) >>>>> >>>>> which gives cc an axistags property. So far however, the algorithm >>>>> always returns 0 synapses. >>>>> >>>>> Also, we have another problem... when we go to visualize our selected >>>>> synapses (from the CC Result overlay, not CC Filtered), we get an error >>>>> that it cannot find the import for: >>>>> >>>>> Loading vtk ... >>>>> Traceback (most recent call last): >>>>> File >>>>> "/home/gt/Desktop/ilastik-0.5.****04-sd_fix/ilastik/gui/** >>>>> >>>>> overlayWidget.py", >>>>> line 206, in onContext >>>>> from mayaviWidget import MayaviQWidget >>>>> File >>>>> "/home/gt/Desktop/ilastik-0.5.****04-sd_fix/ilastik/gui/** >>>>> >>>>> mayaviWidget.py", >>>>> line 6, in >>>>> from enthought.mayavi.core.api import PipelineBase >>>>> ImportError: No module named mayavi.core.api >>>>> >>>>> I've tried putting different .egg folders into the site-packages, but >>>>> no luck. >>>>> >>>>> Thanks for any help. >>>>> ______________________________****_________________ >>>>> ilastik-user mailing list >>>>> ilastik-user at ilastik.org >>>>> http://mailman.ilastik.org/****cgi-bin/mailman/listinfo/**** >>>>> ilastik-user >>>>> >>>>> > >>>>> >>>>> >>>> >>> >>> >> > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at bc.biol.ethz.ch Thu Aug 9 18:27:00 2012 From: christoph.sommer at bc.biol.ethz.ch (Christoph Sommer) Date: Thu, 9 Aug 2012 18:27:00 +0200 Subject: [ilastik-user] ilastik 32-bit version on winXP Message-ID: <5023E4D4.4040301@bc.biol.ethz.ch> Dear all, recently there was a report by Annika Kamper that the currently latest Windows 32-bit version of ilastik is not starting up and just leaves with a DLL import error on Windows XP (on Win 7 it's running). The latest (for win 32-bit) installer is: > ilastik-0.5-setup-v0.5.06.rc5-x86-32.exe Installing the correct Microsoft runtime environments did not help. After some TeamViewer debugging we found the problem and here comes the solution :-) Just delete: \Lib\site-packages\vigra\vigraimpex.dll Cheers, Christoph From mail at moritz-kirschmann.de Fri Aug 10 12:16:55 2012 From: mail at moritz-kirschmann.de (Moritz Kirschmann) Date: Fri, 10 Aug 2012 12:16:55 +0200 Subject: [ilastik-user] ilastik synapse detection Message-ID: <646A1C31-EA48-4E69-91D1-D07E788EAC74@moritz-kirschmann.de> Dear all, I would love to use the ilastik synapse detection on 5nm/pix FIB data. Unfortunately my workstation (24 GB ram) is running win7_64bit. As a workaround I tried to install the ilastik-synapse-0.5.05-mac64.zip on my Macbook (8GB ram, lion). Unfortunately I have troubles running the run-ilastik.sh script: medano:ilastik-synapse-0-1.5.05-mac64 Moritz$ bash run-ilastik.sh dyld: Symbol not found: _objc_release Referenced from: /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation Expected in: /Users/Moritz/Downloads/ilastik-synapse-0-1.5.05-mac64/lib/libobjc.A.dylib in /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation run-ilastik.sh: line 3: 10857 Trace/BPT trap: 5 PYTHONPATH=.:$PYTHONPATH bin/python -B ilastik/ilastikMain.py How can fix this? All help is highly appreciated! I am also available via skype/teamviewer. Thanks in advance! Regards Moritz Kirschmann From anna.kreshuk at iwr.uni-heidelberg.de Fri Aug 10 19:32:32 2012 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Fri, 10 Aug 2012 19:32:32 +0200 Subject: [ilastik-user] ilastik synapse detection In-Reply-To: <646A1C31-EA48-4E69-91D1-D07E788EAC74@moritz-kirschmann.de> References: <646A1C31-EA48-4E69-91D1-D07E788EAC74@moritz-kirschmann.de> Message-ID: <502545B0.80108@iwr.uni-heidelberg.de> Dear Mr. Kirschmann, Thank you for your interest in our synapse detection algorithm. I am not a Mac user myself, so I can't help right now, but next week we'll try to make a new Mac binary for synapse detection, based on the latest version of ilastik. I'll let you know as soon as I have it. Regards Anna Kreshuk On 08/10/2012 12:16 PM, Moritz Kirschmann wrote: > Dear all, > > I would love to use the ilastik synapse detection on 5nm/pix FIB data. Unfortunately my workstation (24 GB ram) is running win7_64bit. As a workaround I tried to install the ilastik-synapse-0.5.05-mac64.zip on my Macbook (8GB ram, lion). > Unfortunately I have troubles running the run-ilastik.sh script: > > medano:ilastik-synapse-0-1.5.05-mac64 Moritz$ bash run-ilastik.sh > dyld: Symbol not found: _objc_release > Referenced from: /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation > Expected in: /Users/Moritz/Downloads/ilastik-synapse-0-1.5.05-mac64/lib/libobjc.A.dylib > in /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation > run-ilastik.sh: line 3: 10857 Trace/BPT trap: 5 PYTHONPATH=.:$PYTHONPATH bin/python -B ilastik/ilastikMain.py > > How can fix this? All help is highly appreciated! I am also available via skype/teamviewer. > Thanks in advance! > > Regards > Moritz Kirschmann > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From hazards at musc.edu Fri Dec 7 22:46:59 2012 From: hazards at musc.edu (Hazard, E. Starr) Date: Fri, 7 Dec 2012 16:46:59 -0500 Subject: [ilastik-user] running on RHEL 6.0 ImportError: No module named ilastik.core Message-ID: <47FF7527B4C3CD42906BFC94B8363D48BC71D7F8EF@EVS5.clinlan.local> Hello I am trying to get ilastik to run (v 0.5.04) I am getting close I think /shared/app/ilastik-0.5.04/bin/python /shared/app/ilastik-0.5.04/ilastik/ilastikMain.py Traceback (most recent call last): File "/shared/app/ilastik-0.5.04/ilastik/ilastikMain.py", line 50, in from ilastik.core import ILASTIK_VERSION ImportError: No module named ilastik.core Where should ilastik core reside? Thanks Starr From thorben.kroeger at iwr.uni-heidelberg.de Mon Dec 10 17:33:38 2012 From: thorben.kroeger at iwr.uni-heidelberg.de (Thorben Kroeger) Date: Mon, 10 Dec 2012 17:33:38 +0100 Subject: [ilastik-user] running on RHEL 6.0 ImportError: No module named ilastik.core In-Reply-To: <47FF7527B4C3CD42906BFC94B8363D48BC71D7F8EF@EVS5.clinlan.local> References: <47FF7527B4C3CD42906BFC94B8363D48BC71D7F8EF@EVS5.clinlan.local> Message-ID: <50C60EE2.2000602@iwr.uni-heidelberg.de> Dear Starr, Try running the run-ilastik-linux.sh script, it will adjust your PYTHONPATH environment variable so that python can find the necessary modules. This file should be bundled with ilastik; you can also view it here: https://github.com/ilastik/ilastik-0.5/blob/master/run-ilastik-linux.sh Ilastik 0.5.04 is quite old, the most recent version is 0.5.12, is there a particular reason why you are using the old version? Good luck, Thorben On 12/07/2012 10:46 PM, Hazard, E. Starr wrote: > Hello > > I am trying to get ilastik to run (v 0.5.04) > > I am getting close I think > > /shared/app/ilastik-0.5.04/bin/python /shared/app/ilastik-0.5.04/ilastik/ilastikMain.py > Traceback (most recent call last): > File "/shared/app/ilastik-0.5.04/ilastik/ilastikMain.py", line 50, in > from ilastik.core import ILASTIK_VERSION > ImportError: No module named ilastik.core > > > Where should ilastik core reside? > > > Thanks > > Starr > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From fred.hamprecht at iwr.uni-heidelberg.de Wed Jul 25 18:51:56 2012 From: fred.hamprecht at iwr.uni-heidelberg.de (Fred Hamprecht) Date: Wed, 25 Jul 2012 18:51:56 +0200 Subject: [ilastik-user] adding previously available labels to a project in ilastik 0.5 In-Reply-To: References: Message-ID: <5010242C.60608@iwr.uni-heidelberg.de> Dear Chao, I think the easiest way is to incorporate the labels in an ilastik project (hf5 file). To see its structure, load the raw data of interest, and annotate it with a few strokes using the mouse. Then save the project as an "ilp" file. This is in hdf5 format which you can browse with any hdf5 viewer. The organization is straightforward, if you load a single image the labels will be stored as a 5D image (with a few singleton dimensions) under DataSets/dataItem00/labels/data. This is where you should insert your own labels using an hdf5 interface which is offered for most high-level languages (though note that matlab seemingly cannot add singleton dimensions to data). Please let me know if that works. Best regards, Fred On 25.07.2012 13:26, Liu, Chao wrote: > Dear Prof. Hamprecht and ilastik team, > [...] > > I would like to use ilastik having the possibility of entering > pre-computed masks for the classes of interest, instead of manually > drawing them. Is it possible to have this functionality in the current > version of ilastik? Or could the ilastik team suggest to me how to do this? > > I would really appreciate any help on this. Looking forward to hearing > from you. > > Best wishes, > Chao Liu From christoph.sommer at bc.biol.ethz.ch Thu Jul 26 09:47:36 2012 From: christoph.sommer at bc.biol.ethz.ch (Christoph Sommer) Date: Thu, 26 Jul 2012 09:47:36 +0200 Subject: [ilastik-user] adding previously available labels to a project in ilastik 0.5 In-Reply-To: <5010242C.60608@iwr.uni-heidelberg.de> References: <5010242C.60608@iwr.uni-heidelberg.de> Message-ID: <5010F618.1090406@bc.biol.ethz.ch> Dear Chao, I attached a small python scaffold script to inject labels (stored in images) into a ilastik project file, to give you an idea how the label data is stored internally. Best, Christoph Am 25.07.2012 18:51, schrieb Fred Hamprecht: > > Dear Chao, > > I think the easiest way is to incorporate the labels in an ilastik > project (hf5 file). To see its structure, load the raw data of interest, > and annotate it with a few strokes using the mouse. Then save the > project as an "ilp" file. This is in hdf5 format which you can browse > with any hdf5 viewer. The organization is straightforward, if you load a > single image the labels will be stored as a 5D image (with a few > singleton dimensions) under DataSets/dataItem00/labels/data. This is > where you should insert your own labels using an hdf5 interface which is > offered for most high-level languages (though note that matlab seemingly > cannot add singleton dimensions to data). > > Please let me know if that works. > > Best regards, > > Fred > > > > > > On 25.07.2012 13:26, Liu, Chao wrote: >> Dear Prof. Hamprecht and ilastik team, >> > [...] >> I would like to use ilastik having the possibility of entering >> pre-computed masks for the classes of interest, instead of manually >> drawing them. Is it possible to have this functionality in the current >> version of ilastik? Or could the ilastik team suggest to me how to do this? >> >> I would really appreciate any help on this. Looking forward to hearing >> from you. >> >> Best wishes, >> Chao Liu > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- import vigra import h5py import numpy # open project file f = h5py.File('project.ilp','a') # get file names of labels label_file_names = ['label_z%05d.png' % i for i in range(50)] # init 5D array (time, x, y, z, channels) labels = numpy.zeros((1,150,100,50,1), numpy.uint8) # fill 5D array for cnt, (r,l) in enumerate(zip(raw_file_names, label_file_names)): lab = vigra.impex.readImage(l) labels[0,:,:, cnt, 0] = lab[:,:,...] # write to hdf5 del f['/DataSetsdataItem00/labels/data'] lg.create_dataset('/DataSetsdataItem00/labels/data', data=labels) f.close() From bernhard.kausler at iwr.uni-heidelberg.de Fri Jul 27 15:53:41 2012 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Fri, 27 Jul 2012 15:53:41 +0200 Subject: [ilastik-user] Viewing and editing labels outside Ilastik with HDFView Message-ID: <50129D65.4070908@iwr.uni-heidelberg.de> Below the protocol of a successfully closed support case regarding viewing and editing labels outside Ilastik with HDFView. User: I would like to use ilastik having the possibility of entering pre-computed masks for the classes of interest, instead of manually drawing them. Is it possible to have this functionality in the current version of ilastik? Or could the ilastik team suggest to me how to do this? Fred Hamprecht: the easiest way is to incorporate the labels in an ilastik project (hf5 file). To see its structure, load the raw data of interest, and annotate it with a few strokes using the mouse. Then save the project as an "ilp" file. This is in hdf5 format which you can browse with any hdf5 viewer. The organization is straightforward, if you load a single image the labels will be stored as a 5D image (with a few singleton dimensions) under DataSets/dataItem00/labels/data. This is where you should insert your own labels using an hdf5 interface which is offered for most high-level languages (though note that matlab seemingly cannot add singleton dimensions to data). User: I created a ilastik project and draw some labels and then save it as a .ilp file. I used the HDFViewer to open it and see its structure, however, all the data in DataSets/dataItem00/labels/data are zero(I can see the information of the labels in properties of label, the attachment is 3 screenshot). Is this what it should be? Christoph Sommer and Bernhard X. Kausler: the labels dataset is five-dimensional (time, 3 spatial dims, channels). Please right-click on the dataset and choose "open as". There you can select the dimensions you want to browse in hdfview. I suggest that you try dim2 for Height and dim3 for Width. I guess that the labels are there but it is hard to see / find them using HDFView. Even if you "Open as image" the labels drawn will be very dark, because the labels are just enumerated. No label has a value zero, label 1 gets value 1, label 2 gets value 2... This means that looking at them in 8-bit gray -cale will show them in almost black. User: Thank you very much for your help. I select the dimension as you suggested and it works! Some of the data are non-zero values and they are exactly the labels I draw From bms2156 at columbia.edu Fri Jul 27 20:12:26 2012 From: bms2156 at columbia.edu (Ben Shababo) Date: Fri, 27 Jul 2012 14:12:26 -0400 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 Message-ID: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> We are trying to run Ilastik 0.5 with Syanpse Detection in Ubuntu 11.04 and have not had much luck. Originally we saw this error when running the bash script run-ilastik.sh: gt at gt-Precision-T7600:~/Desktop/ilastik-0.5.04-sd$ sh run-ilastik.sh Traceback (most recent call last): File "ilastik/ilastikMain.py", line 32, in from OpenGL.GL import * File "/home/gt/Desktop/ilastik-0.5.04-sd/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/GL/__init__.py", line 2, in from OpenGL.raw.GL import * File "/home/gt/Desktop/ilastik-0.5.04-sd/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/__init__.py", line 6, in from OpenGL.raw.GL.constants import * File "/home/gt/Desktop/ilastik-0.5.04-sd/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/constants.py", line 6, in from ctypes import * ImportError: No module named ctypes We've tried many fixes to get around this... we've replaced Python 2.6 with 2.7 in ilastik lib, bin, and include folders. When we do this, the software runs, however, when we press Filter Synapses (following the tutorial and using the sample images from the test data from 'provesynapsedetectionknott') we see this error: Traceback (most recent call last): File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/modules/connected_components/gui/connectedComponentsRibbon.py", line 136, in on_btnFilter_clicked self.parent.project.dataMgr.Connected_Components.filterSynapses(self.inputOverlay, label, minsize, maxsize) File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/modules/connected_components/core/connectedComponentsMgr.py", line 192, in filterSynapses objs_ref = sfad.computeReferenceObjects() File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/modules/connected_components/core/synapseDetectionFilter.py", line 83, in computeReferenceObjects objs_ref = self.objectsSlow3d(cc) File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/modules/connected_components/core/synapseDetectionFilter.py", line 59, in objectsSlow3d nzindex = numpy.nonzero(cc) File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/lib/python2.7/site-packages/numpy/core/fromnumeric.py", line 1176, in nonzero res = nonzero() File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/lib/python2.7/site-packages/vigra/arraytypes.py", line 559, in nonzero res[k].axistags = AxisTags(self.axistags[k]) AttributeError: 'ScalarVolume' object has no attribute 'axistags' We've debugged this a bit, and even created our own ScalarVolume to pass to numpy.nonzero() in the method SynapseDetectionFilter.objectsSlow3d(self, cc) )and this works, but for some reason the ScalarVolume cc which the software passes to numpy.nonzero(cc) returns the error above. Any help would be appreciated. From bms2156 at columbia.edu Fri Jul 27 23:36:51 2012 From: bms2156 at columbia.edu (Ben Shababo) Date: Fri, 27 Jul 2012 17:36:51 -0400 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> Message-ID: <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> Update: We were able to solve the problem when filtering synapses... or at least it doesn't throw an error now... we achieved this by calling cc = ScalarVolume(cc) which gives cc an axistags property. So far however, the algorithm always returns 0 synapses. Also, we have another problem... when we go to visualize our selected synapses (from the CC Result overlay, not CC Filtered), we get an error that it cannot find the import for: Loading vtk ... Traceback (most recent call last): File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/overlayWidget.py", line 206, in onContext from mayaviWidget import MayaviQWidget File "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/mayaviWidget.py", line 6, in from enthought.mayavi.core.api import PipelineBase ImportError: No module named mayavi.core.api I've tried putting different .egg folders into the site-packages, but no luck. Thanks for any help. From anna.kreshuk at iwr.uni-heidelberg.de Sat Jul 28 01:18:18 2012 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sat, 28 Jul 2012 01:18:18 +0200 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> Message-ID: <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> Dear Mr. Shababo, Is this problem very urgent? I think I know where it comes from, but I'm only returning from vacation on Wednesday. Mayavi shouldn't be needed anymore, but the linux binary seems to correspond to an older synapse detection workflow version than the Mac binary. 0 synapses found is suspicious, but it could come from re-initialization of the cc volume. Sorry for the inconvenience Anna Quoting Ben Shababo : > Update: > > We were able to solve the problem when filtering synapses... or at > least it doesn't throw an error now... we achieved this by calling > > cc = ScalarVolume(cc) > > which gives cc an axistags property. So far however, the algorithm > always returns 0 synapses. > > Also, we have another problem... when we go to visualize our selected > synapses (from the CC Result overlay, not CC Filtered), we get an error > that it cannot find the import for: > > Loading vtk ... > Traceback (most recent call last): > File > "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/overlayWidget.py", > line 206, in onContext > from mayaviWidget import MayaviQWidget > File > "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/mayaviWidget.py", > line 6, in > from enthought.mayavi.core.api import PipelineBase > ImportError: No module named mayavi.core.api > > I've tried putting different .egg folders into the site-packages, > but no luck. > > Thanks for any help. > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From bms2156 at columbia.edu Mon Jul 30 15:04:49 2012 From: bms2156 at columbia.edu (Ben Shababo) Date: Mon, 30 Jul 2012 09:04:49 -0400 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> Message-ID: <20120730090449.0dmpqdbwg00k48c0@cubmail.cc.columbia.edu> Anna- Thank you very much for your reply. I'm not exactly sure how urgent the problem is - I'm helping debug though I work in a different group in the lab so I'm not sure about the timeline of the project. In any case, if there is anyone else who could help more in the meantime, that would be beneficial. As far as your comments... the code we are working from definitely imports Mayavi... I can look at the code today to see if it is only imported or if it used as well... and for the 0 synapses issue, my main question is do the axistags in the ScalarVolume matter for your algorithm? As I mentioned, the ScalarVolume cc originally does not have axistags so when we call ScalarVolume(cc), I think it might be the case that the only difference between cc before and after is the addition of the default axistags: x y z. Also, you said you think you know what is causing the issue... would you at least be able to tell me what you think it is and guide me to the relevant sections of code... I can try to fix it in the meantime. Thanks so much! -Ben Quoting anna.kreshuk at iwr.uni-heidelberg.de: > Dear Mr. Shababo, > > Is this problem very urgent? I think I know where it comes from, but > I'm only returning from vacation on Wednesday. Mayavi shouldn't be > needed anymore, but the linux binary seems to correspond to an older > synapse detection workflow version than the Mac binary. 0 synapses > found is suspicious, but it could come from re-initialization of the cc > volume. > > Sorry for the inconvenience > Anna > > > Quoting Ben Shababo : > >> Update: >> >> We were able to solve the problem when filtering synapses... or at >> least it doesn't throw an error now... we achieved this by calling >> >> cc = ScalarVolume(cc) >> >> which gives cc an axistags property. So far however, the algorithm >> always returns 0 synapses. >> >> Also, we have another problem... when we go to visualize our selected >> synapses (from the CC Result overlay, not CC Filtered), we get an error >> that it cannot find the import for: >> >> Loading vtk ... >> Traceback (most recent call last): >> File >> "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/overlayWidget.py", >> line 206, in onContext >> from mayaviWidget import MayaviQWidget >> File >> "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/mayaviWidget.py", >> line 6, in >> from enthought.mayavi.core.api import PipelineBase >> ImportError: No module named mayavi.core.api >> >> I've tried putting different .egg folders into the site-packages, >> but no luck. >> >> Thanks for any help. >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From bms2156 at columbia.edu Mon Jul 30 21:40:22 2012 From: bms2156 at columbia.edu (Ben Shababo) Date: Mon, 30 Jul 2012 15:40:22 -0400 Subject: [ilastik-user] Synapse Detection Download doesn't work with Ubuntu 11.04 In-Reply-To: <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> References: <20120727141226.x128nyb7voskgckg@cubmail.cc.columbia.edu> <20120727173651.zswxlfiae8g80wk4@cubmail.cc.columbia.edu> <20120728011818.dh3z0x5fwcgokock@webmail.iwr.uni-heidelberg.de> Message-ID: <20120730154022.16ovwol9w8gw4g8c@cubmail.cc.columbia.edu> Anna- Also, it may be helpful to know what Linux platform/system was used to develop the Linux version of Ilastik w/ Synapse Detection. Thanks! -Ben Quoting anna.kreshuk at iwr.uni-heidelberg.de: > Dear Mr. Shababo, > > Is this problem very urgent? I think I know where it comes from, but > I'm only returning from vacation on Wednesday. Mayavi shouldn't be > needed anymore, but the linux binary seems to correspond to an older > synapse detection workflow version than the Mac binary. 0 synapses > found is suspicious, but it could come from re-initialization of the cc > volume. > > Sorry for the inconvenience > Anna > > > Quoting Ben Shababo : > >> Update: >> >> We were able to solve the problem when filtering synapses... or at >> least it doesn't throw an error now... we achieved this by calling >> >> cc = ScalarVolume(cc) >> >> which gives cc an axistags property. So far however, the algorithm >> always returns 0 synapses. >> >> Also, we have another problem... when we go to visualize our selected >> synapses (from the CC Result overlay, not CC Filtered), we get an error >> that it cannot find the import for: >> >> Loading vtk ... >> Traceback (most recent call last): >> File >> "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/overlayWidget.py", >> line 206, in onContext >> from mayaviWidget import MayaviQWidget >> File >> "/home/gt/Desktop/ilastik-0.5.04-sd_fix/ilastik/gui/mayaviWidget.py", >> line 6, in >> from enthought.mayavi.core.api import PipelineBase >> ImportError: No module named mayavi.core.api >> >> I've tried putting different .egg folders into the site-packages, >> but no luck. >> >> Thanks for any help. >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From david.kienle at stmail.uni-bayreuth.de Tue Jun 12 16:57:32 2012 From: david.kienle at stmail.uni-bayreuth.de (David Kienle) Date: Tue, 12 Jun 2012 16:57:32 +0200 Subject: [ilastik-user] Ilastik is not running linux ubuntu 12.04 64bit Message-ID: <4FD758DC.2090307@stmail.uni-bayreuth.de> Hello everybody, I tried to run ilastik-0.5-bundle-v0.5.05.tar.gz on an Ubuntu 12.04 (LTS) 64bit system. Unfortunately I get this error message: david at notebook-david:~/Desktop/ilastik-0.5.04$ sudo ./run-ilastik.sh [sudo] password for david: Traceback (most recent call last): File "ilastik/ilastikMain.py", line 32, in from OpenGL.GL import * File "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/GL/__init__.py", line 2, in from OpenGL.raw.GL import * File "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/__init__.py", line 6, in from OpenGL.raw.GL.constants import * File "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/constants.py", line 6, in from ctypes import * ImportError: No module named ctypes Python 2.6.7 is installed and could also be initialized by pythonbrew (python-ctypes is also installed, also installing the newest version of ctypes (direct, not via ubuntu sources) didn't help). I also tried using the sources but qimage2ndarray is today not available in the right way - no way. Did anyone used Ilastik with a new version of Ubuntu? Or are there some strange and uncomfortable dependencies? Thanks for help! David From christoph.sommer at bc.biol.ethz.ch Tue Jun 12 17:04:40 2012 From: christoph.sommer at bc.biol.ethz.ch (Christoph Sommer) Date: Tue, 12 Jun 2012 17:04:40 +0200 Subject: [ilastik-user] Ilastik is not running linux ubuntu 12.04 64bit In-Reply-To: <4FD758DC.2090307@stmail.uni-bayreuth.de> References: <4FD758DC.2090307@stmail.uni-bayreuth.de> Message-ID: <4FD75A88.4060407@bc.biol.ethz.ch> Hi David, thanks for your report! I have a few comments and questions: * Why are running it as root? * What happens if you: python -c "import ctypes"? * In general I would recommend python2.7!! > I also tried using the sources but qimage2ndarray is today not > available in the right way - no way. * Did you try to easy_install / pip qimage2ndarray? All the Best, Christoph I think Christoph Straehle in Heidelberg uses the newest Ubuntu. Am 12.06.2012 16:57, schrieb David Kienle: > Hello everybody, > > I tried to run ilastik-0.5-bundle-v0.5.05.tar.gz on an Ubuntu 12.04 > (LTS) 64bit system. Unfortunately I get this error message: > > david at notebook-david:~/Desktop/ilastik-0.5.04$ sudo ./run-ilastik.sh > [sudo] password for david: > Traceback (most recent call last): > File "ilastik/ilastikMain.py", line 32, in > from OpenGL.GL import * > File > "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/GL/__init__.py", > line 2, in > from OpenGL.raw.GL import * > File > "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/__init__.py", > line 6, in > from OpenGL.raw.GL.constants import * > File > "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/constants.py", > line 6, in > from ctypes import * > ImportError: No module named ctypes > > Python 2.6.7 is installed and could also be initialized by pythonbrew > (python-ctypes is also installed, also installing the newest version > of ctypes (direct, not via ubuntu sources) didn't help). > > I also tried using the sources but qimage2ndarray is today not > available in the right way - no way. > > Did anyone used Ilastik with a new version of Ubuntu? Or are there > some strange and uncomfortable dependencies? > > Thanks for help! > > David > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From david.kienle at stmail.uni-bayreuth.de Mon Jun 18 16:19:01 2012 From: david.kienle at stmail.uni-bayreuth.de (David Kienle) Date: Mon, 18 Jun 2012 16:19:01 +0200 Subject: [ilastik-user] Ilastik is not running linux ubuntu 12.04 64bit In-Reply-To: <4FD75A88.4060407@bc.biol.ethz.ch> References: <4FD758DC.2090307@stmail.uni-bayreuth.de> <4FD75A88.4060407@bc.biol.ethz.ch> Message-ID: <4FDF38D5.5040603@stmail.uni-bayreuth.de> Hi Christoph, thanks for your answer and help. > * Why are running it as root? That's awkward - I don't know why I did this as root but that was not necessary. > * What happens if you: python -c "import ctypes"? Import is working, but nothing changes. Also tried to use the original "ctypes" (not from the Ubuntu sources). > * In general I would recommend python2.7!! I also used python2.7 but cause of the explanations in the Ilastik installation explanations (http://ilastik.org/index.php?cat=15_Download&page=06_Installation) I also tried this. > * Did you try to easy_install / pip qimage2ndarray? Yes, I tried. But now qimage2ndarray is working - it was deppending on other moduls I've supplemented. Now I also tried to install the source version from the git repository. But this error message is mysterious: david at notebook-david:~/Desktop/Ilastik-ilastik-fbcb105$ ./run-ilastik.sh Traceback (most recent call last): File "/home/david/Desktop/Ilastik-ilastik-fbcb105/ilastik/ilastikMain.py", line 30, in import vigra File "/usr/lib/python2.7/dist-packages/vigra/__init__.py", line 101, in import vigranumpycore ImportError: libboost_python.so.1.49.0: cannot open shared object file: No such file or directory vigra 1.8.0 is already well installed, also libboost_python.so.1.49.0 is in /usr/local/lib/ (looks good). Setting LD_LIBRARY_PATH=/usr/local/lib/ didn't change it - but maybe that's not the right way... Thanks and all the best, David > > All the Best, > Christoph > > > I think Christoph Straehle in Heidelberg uses the newest Ubuntu. > > Am 12.06.2012 16:57, schrieb David Kienle: >> Hello everybody, >> >> I tried to run ilastik-0.5-bundle-v0.5.05.tar.gz on an Ubuntu >> 12.04 (LTS) 64bit system. Unfortunately I get this error message: >> >> david at notebook-david:~/Desktop/ilastik-0.5.04$ sudo ./run-ilastik.sh >> [sudo] password for david: >> Traceback (most recent call last): >> File "ilastik/ilastikMain.py", line 32, in >> from OpenGL.GL import * >> File >> "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/GL/__init__.py", >> line 2, in >> from OpenGL.raw.GL import * >> File >> "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/__init__.py", >> line 6, in >> from OpenGL.raw.GL.constants import * >> File >> "/home/david/Desktop/ilastik-0.5.04/lib/python2.6/site-packages/PyOpenGL-3.0.1-py2.6.egg/OpenGL/raw/GL/constants.py", >> line 6, in >> from ctypes import * >> ImportError: No module named ctypes >> >> Python 2.6.7 is installed and could also be initialized by >> pythonbrew (python-ctypes is also installed, also installing the >> newest version of ctypes (direct, not via ubuntu sources) didn't >> help). >> >> I also tried using the sources but qimage2ndarray is today not >> available in the right way - no way. >> >> Did anyone used Ilastik with a new version of Ubuntu? Or are there >> some strange and uncomfortable dependencies? >> >> Thanks for help! >> >> David >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > From jason.maynes at sickkids.ca Wed Jun 20 01:24:59 2012 From: jason.maynes at sickkids.ca (Jason Maynes) Date: Tue, 19 Jun 2012 23:24:59 +0000 Subject: [ilastik-user] cell profiler classify pixels problem Message-ID: <682E232A-B8AF-4F59-9832-010C4A0712DC@sickkids.ca> I have a very nice classification that ilastik generated but when I attempt to use it in cell profiler, the segmentation is very different (on one of the images that was used to generate the classifier even). I am doing everything as described on the ilastik website for cell profiler use (not exactly a lot of options to tweak so pretty straight forward). Any ideas as to why the classify pixels comes up with such a different result in cell profiler? Thanks. Cheers, JTM **************************************** "...everything that living things do can be understood in terms of the jigglings and wigglings of atoms." Richard Feynman, Nobel Laureate 1965 "We can be sure that if a detailed understanding of the molecular basis of chemo-therapeutic activity were to be obtained, the advance of medicine would be greatly accelerated." Linus Pauling, Nobel Laureate 1954 **************************************** Jason Thomas Maynes, PhD/MD Departments of Paediatric Anaesthesia and Molecular Structure and Function Hospital for Sick Children/SickKids Research Institute Faculty of Medicine University of Toronto Office: 416-813-5934 Cell: 647-746-6198 jason.maynes at sickkids.ca **************************************** ________________________________ This e-mail may contain confidential, personal and/or health information(information which may be subject to legal restrictions on use, retention and/or disclosure) for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies. From p.van.loon at rijkzwaan.nl Wed Jun 20 12:32:34 2012 From: p.van.loon at rijkzwaan.nl (Peter van Loon) Date: Wed, 20 Jun 2012 12:32:34 +0200 Subject: [ilastik-user] cell profiler classify pixels problem In-Reply-To: References: Message-ID: Hi Jason, I had this once when I was using 32 bit RGB images in Ilastik. It seemed to work alright but in CellProfiler the results looked strange. Now I use only 24 bit images in Ilastik and CellProfiler. Regards, Peter van Loon From: ilastik-user-request at ilastik.org To: ilastik-user at ilastik.org Date: 20-06-2012 12:00 Subject: ilastik-user Digest, Vol 2, Issue 5 Sent by: ilastik-user-bounces at ilastik.org Send ilastik-user mailing list submissions to ilastik-user at ilastik.org To subscribe or unsubscribe via the World Wide Web, visit http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user or, via email, send a message with subject or body 'help' to ilastik-user-request at ilastik.org You can reach the person managing the list at ilastik-user-owner at ilastik.org When replying, please edit your Subject line so it is more specific than "Re: Contents of ilastik-user digest..." Today's Topics: 1. cell profiler classify pixels problem (Jason Maynes) ---------------------------------------------------------------------- Message: 1 Date: Tue, 19 Jun 2012 23:24:59 +0000 From: Jason Maynes To: "ilastik-user at ilastik.org" Subject: [ilastik-user] cell profiler classify pixels problem Message-ID: <682E232A-B8AF-4F59-9832-010C4A0712DC at sickkids.ca> Content-Type: text/plain; charset="us-ascii" I have a very nice classification that ilastik generated but when I attempt to use it in cell profiler, the segmentation is very different (on one of the images that was used to generate the classifier even). I am doing everything as described on the ilastik website for cell profiler use (not exactly a lot of options to tweak so pretty straight forward). Any ideas as to why the classify pixels comes up with such a different result in cell profiler? Thanks. Cheers, JTM **************************************** "...everything that living things do can be understood in terms of the jigglings and wigglings of atoms." Richard Feynman, Nobel Laureate 1965 "We can be sure that if a detailed understanding of the molecular basis of chemo-therapeutic activity were to be obtained, the advance of medicine would be greatly accelerated." Linus Pauling, Nobel Laureate 1954 **************************************** Jason Thomas Maynes, PhD/MD Departments of Paediatric Anaesthesia and Molecular Structure and Function Hospital for Sick Children/SickKids Research Institute Faculty of Medicine University of Toronto Office: 416-813-5934 Cell: 647-746-6198 jason.maynes at sickkids.ca **************************************** ________________________________ This e-mail may contain confidential, personal and/or health information(information which may be subject to legal restrictions on use, retention and/or disclosure) for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies. ------------------------------ _______________________________________________ ilastik-user mailing list ilastik-user at ilastik.org http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user End of ilastik-user Digest, Vol 2, Issue 5 ****************************************** The information contained in this e-mail (and attachments if any) is exclusively intended for the recipient(s) named above and may be confidential, proprietary, and/or legally privileged. Unintentional disclosure of this e-mail does not constitute a waiver of any right or privilege. Please notify us if you are not the intended recipient and do not use, print, copy, forward, or disclose (any part of) this e-mail, but delete it (and attachments and copies if any) subsequently. Thank you. You can find disclaimer translations in different languages by visiting our website http://www.rijkzwaan.com/edisclaimer The registration number of each Rijk Zwaan company can be found by using the following link: http://www.rijkzwaan.com/registrationnumbers -------------- next part -------------- An HTML attachment was scrubbed... URL: From Paul.Buscemi at wuxiapptec.com Tue Jun 26 14:59:58 2012 From: Paul.Buscemi at wuxiapptec.com (Buscemi, Paul) Date: Tue, 26 Jun 2012 07:59:58 -0500 Subject: [ilastik-user] startup Message-ID: <505680C0D921E443A9FAB3EAE458F09F0609C1FB25@E2K7CCR.US.pharmatechs.com> I can run Ilastik .5.05 on a win 7 32 bit machine, but .5.05- rcs will not start. The cmd screen opens but closes too fast for me to read, but it appears only a few lines are written. Suggestions ? Thanks... Paul Buscemi, Ph.D. Senior Director Study Development Wuxi Apptec, Inc. 2540 Executive Drive, St Paul MN, 55120 Paul.buscemi at wuxiapptec.com 763 567 1379 c 651 675 2077 o ________________________________ Save paper. Protect the environment. Print only when necessary. This e-mail transmission may contain confidential or legally privileged information belonging to WuXi AppTec Co., Ltd. The content of this transmission is intended only for the individual or entity named in the e-mail address. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution, or reliance upon the content of this e-mail is strictly prohibited. If you have received this e-mail transmission in error, please reply to sender so that we can arrange correct delivery. Then please delete this message from your inbox. -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at bc.biol.ethz.ch Tue Jun 26 15:14:10 2012 From: christoph.sommer at bc.biol.ethz.ch (Christoph Sommer) Date: Tue, 26 Jun 2012 15:14:10 +0200 Subject: [ilastik-user] startup In-Reply-To: <505680C0D921E443A9FAB3EAE458F09F0609C1FB25@E2K7CCR.US.pharmatechs.com> References: <505680C0D921E443A9FAB3EAE458F09F0609C1FB25@E2K7CCR.US.pharmatechs.com> Message-ID: <4FE9B5A2.5040807@bc.biol.ethz.ch> Dear Paul, can you open a command shell, cd to the ilastik installation path and run "run_ilastik.bat". This should give the error message. Did you install the Microsoft runtime libraries? Do you have access to Win 7 64 bit machine? In future the windows binaries will be 64 bit. Best, Christoph Am 26.06.2012 14:59, schrieb Buscemi, Paul: > > I can run Ilastik .5.05 on a win 7 32 bit machine, but .5.05- rcs > will not start. The cmd screen opens but closes too fast for me to > read, but it appears only a few lines are written. > > Suggestions ? > > Thanks... > > *Paul Buscemi, Ph.D.* > > Senior Director Study Development > > Wuxi Apptec, Inc. > > 2540 Executive Drive, St Paul MN, 55120 > > Paul.buscemi at wuxiapptec.com > > 763 567 1379 c 651 675 2077 o > > > ------------------------------------------------------------------------ > Save paper. Protect the environment. Print only when necessary. > > This e-mail transmission may contain confidential or legally > privileged information belonging to WuXi AppTec Co., Ltd. The content > of this transmission is intended only for the individual or entity > named in the e-mail address. If you are not the intended recipient, > you are hereby notified that any disclosure, copying, distribution, or > reliance upon the content of this e-mail is strictly prohibited. > > If you have received this e-mail transmission in error, please reply > to sender so that we can arrange correct delivery. Then please delete > this message from your inbox. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From bruno.cornelis at gmail.com Thu Sep 20 05:04:52 2012 From: bruno.cornelis at gmail.com (Bruno Cornelis) Date: Wed, 19 Sep 2012 23:04:52 -0400 Subject: [ilastik-user] installation of ilastik on Mac OS X Lion (10.7.4) Message-ID: I have been trying to install Ilastik (v0.5.12) from source on Mac OS X Lion but I never seem to be able to succeed.I have Xcode 4.3.2 and Xcode 4.2 developer installed. I have tried altering the "install-ilastik-deps.py" script by changing some lines but still get many errors (mainly compile issues). Is there a newer version of this script that is compatible with later versions of OS X Lion? I did change the following lines to the script: __builtin__.installDir = os.environ["HOME"] --> __builtin__.installDir = '/Users/bcorneli/ilastik' os.environ["MACOSX_DEPLOYMENT_TARGET"]="10.6" --> os.environ["MACOSX_DEPLOYMENT_TARGET"]="10.7" __builtin__.pythonVersionPath = installDir+"/Frameworks/Python.framework/Versions/"+pythonVersion --> __builtin__.pythonVersionPath = "/opt/local/Library/Frameworks/Python.framework/Versions/"+pythonVersion I am not sure whether I am supposed to change this last line or leave it as is. Does this point to my current Python installation or the one being installed as a dependency? But even when I leave it as is, I still get many errors. Also I tried installing the dependencies myself, but even then I get errors when trying to run for example this script "getImportantVariables.py" Would switching to 0.6 by any chance be easier? Any pointers or help are greatly appreciated. Thanks! -------------- next part -------------- An HTML attachment was scrubbed... URL: From bernhard.kausler at iwr.uni-heidelberg.de Tue Apr 2 11:11:23 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Tue, 02 Apr 2013 11:11:23 +0200 Subject: [ilastik-user] [ilastik-announce] Z stack Loading In-Reply-To: References: Message-ID: <515AA0BB.904@iwr.uni-heidelberg.de> Dear Aura, please make sure that you followed the instructions to load a 3d stack closely: http://www.ilastik.org/index.php?cat=20_Documentation&page=02_Data%20loading#load_3d_stack In particular, one should not use the "Add" button but the "Load stack" button. If you like, please send me a small sample of your data (3 images are enough): bernhard.kausler at iwr.uni-heidelberg.de Then I could investigate if there are any issues with your data. Best, Bernhard On 04/02/2013 10:21 AM, Putri Aura wrote: > Dear all, > I am new in using ilastik. After watching the video about the 3D > classification, I wanted to load a 3D EM 8bit-tif z-stack but only > got a single image on screen. The stack contains 100 slices > (1024x1024x100) with 6 nm pixel size for x and y, 30 nm for voxel > depth. The 3d stack from the demo folder could display as a 3d stack. I > have no idea how to load the *.tif stack. Please someone > could help me with this question. I appreciate your any assistance. > Best regards, > aura > > > _______________________________________________ > ilastik-announce mailing list > ilastik-announce at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-announce > From salem at exchange.nih.gov Tue Apr 16 21:35:15 2013 From: salem at exchange.nih.gov (Salem, Ghadi (NIH/CIT) [E]) Date: Tue, 16 Apr 2013 19:35:15 +0000 Subject: [ilastik-user] problem starting on windows 7 Message-ID: <111C1E43DB683D46A1D382DE1909B394068B5CA3@MLBXv02.nih.gov> Hi all, I installed Ilastik on my Windows 7 (64-bit) machine. I have the 2008 and 2010 redistribution packages in place. But, when I start a cmd window to run ilastik, I get the following: C:\Program Files\ilastik-0.5.12>run-ilastik Traceback (most recent call last): File "ilastik\ilastikMain.py", line 30, in import vigra ImportError: No module named vigra 'sleep' is not recognized as an internal or external command, operable program or batch file. When I check the path, only the dlls of ilastik shows up (see lines below). C:\Program Files\ilastik-0.5.12>path PATH=C:\Program Files\ilastik-0.5.12\dlls; [... etc] I've added the whole ilastik directory to the path, and that didn't help. This looks like a powerful software, I'm anxious to use it. Please help. Ghadi -------------- next part -------------- An HTML attachment was scrubbed... URL: From martin.o'reilly at ucl.ac.uk Wed Apr 24 20:04:05 2013 From: martin.o'reilly at ucl.ac.uk (O'Reilly, Martin) Date: Wed, 24 Apr 2013 18:04:05 +0000 Subject: [ilastik-user] Using saved classifier for batch processing Message-ID: Hi, I would like to use a previously saved classifier to batch process some additional data. I can't find a way to load the saved classifier .h5 file from the GUI. Am I missing something obvious here? If not, is there a way to do this by calling the underlying batch processing code from a python script? Thanks in advance, Martin -------------- next part -------------- An HTML attachment was scrubbed... URL: From bernhard.kausler at iwr.uni-heidelberg.de Thu Apr 25 23:00:31 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Thu, 25 Apr 2013 23:00:31 +0200 Subject: [ilastik-user] Using saved classifier for batch processing In-Reply-To: References: Message-ID: <5179996F.2030305@iwr.uni-heidelberg.de> Dear Martin, unfortunately, there is no such function. However, there is workaround. Ilastik project files (".ilp") are just hdf5 files (as you can check for yourself by opening them with hdfview). The classifier is stored in the group "Project/classifiers" inside the project file and it is exactly this group that is exported when using the "export classifier" functionality. So, make a new ilastik project with exactly the same features as you selected before, delete the "classifiers" group and copy in the data from your exported classifier file. Regarding calling ilastik from a Python script: this can be done easily. Have a look inside the "ilastik" subfolder in your ilastik installation. It is just a Python library which can be imported in your scripts if the installation dir is appending to the PYTHONPATH. Hope that that solves your problem. Best, Bernhard On 04/24/2013 08:04 PM, O'Reilly, Martin wrote: > Hi, > > I would like to use a previously saved classifier to batch process some > additional data. I can't find a way to load the saved classifier .h5 > file from the GUI. Am I missing something obvious here? If not, is there > a way to do this by calling the underlying batch processing code from a > python script? > > Thanks in advance, > > Martin > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From bernhard.kausler at iwr.uni-heidelberg.de Thu Apr 25 23:10:49 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Thu, 25 Apr 2013 23:10:49 +0200 Subject: [ilastik-user] problem starting on windows 7 In-Reply-To: <111C1E43DB683D46A1D382DE1909B394068B5CA3@MLBXv02.nih.gov> References: <111C1E43DB683D46A1D382DE1909B394068B5CA3@MLBXv02.nih.gov> Message-ID: <51799BD9.8050907@iwr.uni-heidelberg.de> Dear Ghadi, we tested ilastik 0.5 on Windows 7 64bit and it should work. I see that you tried to start it from a command prompt. Please download the installer exe for Win7 64 from http://www.ilastik.org/index.php?cat=Download and try to start ilastik from the start menu entry. That should avoid any problems with import paths. Best, Bernhard On 04/16/2013 09:35 PM, Salem, Ghadi (NIH/CIT) [E] wrote: > Hi all, > > I installed Ilastik on my Windows 7 (64-bit) machine. I have the 2008 > and 2010 redistribution packages in place. But, when I start a cmd > window to run ilastik, I get the following: > > C:\Program Files\ilastik-0.5.12>run-ilastik > > Traceback (most recent call last): > > File "ilastik\ilastikMain.py", line 30, in > > import vigra > > ImportError: No module named vigra > > 'sleep' is not recognized as an internal or external command, > > operable program or batch file. > > When I check the path, only the dlls of ilastik shows up (see lines below). > > C:\Program Files\ilastik-0.5.12>path > > PATH=C:\Program Files\ilastik-0.5.12\dlls; [? etc] > > I?ve added the whole ilastik directory to the path, and that didn?t help. > > This looks like a powerful software, I?m anxious to use it. Please help. > > Ghadi > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From darrenslatten at gmail.com Tue Apr 30 00:33:48 2013 From: darrenslatten at gmail.com (Darren Slatten) Date: Mon, 29 Apr 2013 17:33:48 -0500 Subject: [ilastik-user] Feasibility of Real-Time Video Stream Processing Message-ID: Hello, I have a rather unique project that I'm working on, and I'm trying to determine if ilastik would be a good fit. I don't have any experience with image processing, so ilastik's intuitive interface is definitely a huge plus. Anyway, here's an overview of what I'm trying to accomplish... In a nutshell: I'm trying to build a robot that can play video games. (Yes, I realize how crazy that sounds. :) The system, as I currently envision it, will consist of the following components: - Video game console - receives instructions from video game controller buttons - sends digital video signal to PC - PC - monitor - displays video game - media software - receives video signal - converts video signal into image stream (i.e., individual frames) - sends image stream to image processor - image processor - receives image stream - checks each frame for a predefined set of features - based on the features, calculates a set of property values for each *frame object* - [simple example] Does the image contain the enemy1 pixel/feature signature ? frame.enemy1.exists = true : frame.enemy1.exists = false; - [medium example] If frame.enemy1.exists, set frame.enemy1.position.x and frame.enemy1.position.y - [advanced example] If frame.enemy1.exists and previousFrame.enemy1.exists, set frame.enemy1.direction and frame.enemy1.velocity - sends frame object stream to decision engine - decision engine - receives frame object stream from image processor - applies predefined rules for translating frame objects into instructions (for the robot) - sends instructions to bluetooth transmitter - bluetooth transmitter - receives instructions from decision engine - sends instructions to robot's bluetooth receiver - The robot - bluetooth receiver - receives instructions from bluetooth transmitter - sends instructions to central processing unit - central processing unit - receives instructions from bluetooth receiver - has wired connections to stepper motors and servos in robot's fingers - based on instructions, activates the appropriate stepper motors and servos, causing robot's fingers to press (or release) video game controller buttons - fingers - uses stepper motors and servos to physically press or release buttons on the video game controller Basically, the PC will handle all the "heavy lifting" and send simple instructions to the robot via bluetooth. The end goal is to have a robot that's capable of playing video games by reacting to specific features of each frame, *in real time*. I'm fairly confident that I can program and connect each of these components, but the piece that loses me is the *PC >> image processor*. More generally, I don't know how to (1) define/codify image features, (2) test each image for those features, and (3) translate the test results into data objects that the decision engine can consume. For relatively simple video games--let's say Pong, as an example--there would be a small number of image features to codify. But as the video games increase in complexity, and the number of image features has to scale up to accommodate every possible scenario in a given game, the process of defining and testing new image features becomes the most time-consuming part of this project. Therefore, since ilastik's user interface seems especially well-suited for handling that task efficiently, I'm willing to design the entire system around it. But before I invest the time and energy required to do that, I'd appreciate some feedback from someone who's familiar with ilastik (and image processing in general). Specifically, I'd like to know if ilastik is a viable option for my project, and if not, what custom components, extensions, etc. would I need to develop to make it work? So without further ado, here are some specific questions/issues that I'm hoping someone can help me with... 1. Can ilastik process images directly from a video signal? 2. If not, can ilastik process images from an image stream? 3. If not, is it possible to integrate ilastik in a script that could, e.g., initiate a feature test for every image file that's added to a given directory? 4. If ilastik is called from a custom script, what kinds of information can it return? For example, if I want to check for a certain pixel pattern, can ilastik return the x-y coordinates of each instance it finds? 5. Do you have any sample data that shows approximately how long it takes ilastik to process a single image (e.g., for a given hardware configuration, image size, feature set, etc.)? 6. Considering the fact that my images and features can be defined with extreme confidence and precision (e.g., the ball in Pong is always the same exact number of pixels, in the same color, in the same layout), is there an image processing tool/library that is better suited for this application? I know this is a lot of information to digest, but any advice you can offer me would be greatly appreciated. I'm not really looking for someone to explain *how* to go about this project--I'm more interested in knowing *if*it's possible with ilastik. And if not...maybe just point me in the right direction. Thanks for your time. :) -- Darren Slatten -------------- next part -------------- An HTML attachment was scrubbed... URL: From berman.albert at gmail.com Sun Aug 4 01:00:19 2013 From: berman.albert at gmail.com (Albert Berman) Date: Sun, 4 Aug 2013 02:00:19 +0300 Subject: [ilastik-user] how-to learn the boundary weights between different objects. Message-ID: Dear ilastik-user, I am new to the ilastik community, and may be asking a well known question: How to learn/taught ilastik to differentiate the boundaries between object and background the best? Is there a special strategy for boundaries ? except of putting labels to the object and to the background it self? thank you in advance, Albert -------------- next part -------------- An HTML attachment was scrubbed... URL: From cyril.mongis at googlemail.com Tue Aug 27 12:36:27 2013 From: cyril.mongis at googlemail.com (Mongis Cyril) Date: Tue, 27 Aug 2013 12:36:27 +0200 Subject: [ilastik-user] iLastik in command line Message-ID: Dear Sirs, I'm very impressed by the work that has been done on iLastik graphical interface in the past months, as much as I'm impressed by its segmentation capabilities. However, as a "power-user", I would like to use iLastik capabilities in a automated workflow. It means that I would like to save the training done before with some pictures (using the graphical interface) and call this training in command line, applying it to some images. Is there a way to do it in iLastik ? I use the 64-bit Linux version. Cheers, Cyril MONGIS KnopLab, ZMBH, Heidelberg http://www.cyrilmongis.net -------------- next part -------------- An HTML attachment was scrubbed... URL: From jgburchfield.au at gmail.com Mon Dec 2 09:33:37 2013 From: jgburchfield.au at gmail.com (James Burchfield) Date: Mon, 2 Dec 2013 19:33:37 +1100 Subject: [ilastik-user] ilastik-user Digest, Vol 15, Issue 4 In-Reply-To: References: Message-ID: Hi Dana I would suggest you use either Cell Profiler or image j It would be pretty straight forward to turn your segments into volume measurements James On 28/11/2013 10:00 PM, wrote: > Send ilastik-user mailing list submissions to > ilastik-user at ilastik.org > > To subscribe or unsubscribe via the World Wide Web, visit > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > or, via email, send a message with subject or body 'help' to > ilastik-user-request at ilastik.org > > You can reach the person managing the list at > ilastik-user-owner at ilastik.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of ilastik-user digest..." > > > Today's Topics: > > 1. Re: a question regarding Ilastik (Anna Kreshuk) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Wed, 27 Nov 2013 14:33:19 +0100 > From: Anna Kreshuk > To: dafi sourani , ilastik-user at ilastik.org > Subject: Re: [ilastik-user] a question regarding Ilastik > Message-ID: <5295F49F.8060100 at iwr.uni-heidelberg.de> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Dear Daphna, > > I'm glad to hear you are finding ilastik so useful for you. > Unfortunately, we don't have a way to do volumetric measurements. To > export the resulting segmentation and measure the objects in another > program, try to use the "write segmentation" option in the automated > classification tab. This will save the segmentation images rather than > probability maps. There are also ways to save individual objects from > seeded watershed (carving), but other people on the team are more > familiar with them than I am. I've cc-ed my reply to our mailing list > and I'll ask them to answer. > > Best, > Anna > > On 27.11.2013 13:39, dafi sourani wrote: > > Dea Dr. Kreshuk, > > My name is Daphna, a PhD student for medical science using MRI. > > I started using Ilastik with my MRI data about two weeks ago and I > > found it very useful, friendly and helpful. > > First, thank you for developng such a great program. > > If possible, I would like to ask a question: My goal is to segment a > > few importants organs and to calculate their volumes. > > For the segmentation I used both automate classification and seeded > > watershed (for different purposes). However, I wonder if there is a > > tool which calculates the total volume of the segmented organs. > > If not- how can I export the coordinates of the pixels containing the > > volume in order to calculate the volume in Matlab or other programming > > tool? > > Thank you very much in advence, > > All the best, > > Daphna > > > > ------------------------------ > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > End of ilastik-user Digest, Vol 15, Issue 4 > ******************************************* > -------------- next part -------------- An HTML attachment was scrubbed... URL: From bohumil.maco at epfl.ch Wed Dec 4 09:49:34 2013 From: bohumil.maco at epfl.ch (Maco Bohumil) Date: Wed, 4 Dec 2013 08:49:34 +0000 Subject: [ilastik-user] segmentation overlay h5 file Message-ID: Hi developer(s), it is possible to merge multiple h5 files of the segmentation overlay to the single one? I am working on a project where we are reconstructing all the objects in our large volume of FIBSEM image stacks. At the end of the session I save the overlay segmentation as h5 file in order to keep track of the object already segmented. When I start new session in ilastik I import that segmentation overlay from h5 file. As the project is growing I have many h5 files and loading them one by one is not very convenient also ending up with long list to scroll. So I was wondering if is possible to merge those h5 files into a single one (all are from the same volume e.g. the same dimensions). And the second question: Sometimes I was in a hurry not to change the color of seeds to the same black color for all the seeds and during importing none of the option (use color from the file, use grey or use selected color (as black)) works. It is possible to change seeds color somehow editing the h5 file? Thank you very much in advance and I am looking forward for your suggestions. Cheers, Bohumil Maco -------------- next part -------------- An HTML attachment was scrubbed... URL: From mahesh.vk at uni-jena.de Wed Dec 4 12:50:18 2013 From: mahesh.vk at uni-jena.de (Mahesh Venkata Krishna) Date: Wed, 4 Dec 2013 12:50:18 +0100 Subject: [ilastik-user] Compilation problems Message-ID: Hi everyone, I am trying to compile ilastik-0.5 on openSuSe 12.1. Since it required vigra module to run, I cloned the vigra-ilastik-0.5 from github and tried compiling with the option "-DWITH_VIGRANUMPY=1". I get the following error when I try to run "make": Scanning dependencies of target vigranumpy_core [ 52%] Building CXX object vigranumpy/src/core/CMakeFiles/vigranumpy_core.dir/vigranumpycore.cxx.o In file included from /home/mahesh/workspace/vigra-ilastik-05/vigranumpy/src/core/vigranumpycore.cxx:41:0: /home/mahesh/workspace/vigra-ilastik-05/include/vigra/numpy_array.hxx:50:31: fatal error: numpy/arrayobject.h: Datei oder Verzeichnis nicht gefunden compilation terminated. make[2]: *** [vigranumpy/src/core/CMakeFiles/vigranumpy_core.dir/vigranumpycore.cxx.o] Fehler 1 make[1]: *** [vigranumpy/src/core/CMakeFiles/vigranumpy_core.dir/all] Fehler 2 make: *** [all] Fehler 2 How can I fix it? Thanks in advance, Mahesh -------------- next part -------------- An HTML attachment was scrubbed... URL: From thorben.kroeger at iwr.uni-heidelberg.de Wed Dec 4 17:48:10 2013 From: thorben.kroeger at iwr.uni-heidelberg.de (Thorben Kroeger) Date: Wed, 04 Dec 2013 17:48:10 +0100 Subject: [ilastik-user] segmentation overlay h5 file In-Reply-To: References: Message-ID: <529F5CCA.5090200@iwr.uni-heidelberg.de> Hello Bohumil, On 12/04/2013 09:49 AM, Maco Bohumil wrote: > it is possible to merge multiple h5 files of the segmentation overlay to > the single one? You are working with ilastik version 0.5 and the carving workflow, correct? > I am working on a project where we are reconstructing all the objects in > our large volume of FIBSEM image stacks. At the end of the session I > save the overlay segmentation as h5 file in order to keep track of the > object already segmented. So you are segmenting one object at a time? There are two issues: 1) the segmentation volume is not compressed, thus significant storage space is needed 2) objects may overlap (as they are extracted independently) In the new version of ilastik, 1.0 (download at http://ilastik.github.io/index10.html) we have re-written carving to support the workflow you are currently using. Each carved object can be saved under a name; it will be stored compressed. ilastik can show all completed objects, they can be loaded again for correctsions and so on. Please have a look at http://ilastik.github.io/documentation/carving/carving.html for the documentation of carving in version 1.0 When I start new session in ilastik I import > that segmentation overlay from h5 file. As the project is growing I have > many h5 files and loading them one by one is not very convenient also > ending up with long list to scroll. > > So I was wondering if is possible to merge those h5 files into a single > one (all are from the same volume e.g. the same dimensions). The best way (other than switching to the new version) would be to make a colored volume image. For each carved object, pick a random color. Then write an HDF5 file (dimensions (1,#x,#y,#z,3) with a RGB color for each x,y,z voxel. The remaining problem is what to do when your objects overlap, but this is a good check anyway in order to see how precise your carving is (if the overlap is significant, the tracer likely made a mistake like missing a branch). > > And the second question: Sometimes I was in a hurry not to change the > color of seeds to the same black color for all the seeds and during > importing none of the option (use color from the file, use grey or use > selected color (as black)) works. It is possible to change seeds color > somehow editing the h5 file? I'm not sure I understand. If you forgot to change the seed color, and then you load such a carved object back as .h5 file, the object appears in another color? Why does it matter what color the object has? All the best, Thorben From thorben.kroeger at iwr.uni-heidelberg.de Thu Dec 5 08:27:25 2013 From: thorben.kroeger at iwr.uni-heidelberg.de (Thorben Kroeger) Date: Thu, 05 Dec 2013 08:27:25 +0100 Subject: [ilastik-user] segmentation overlay h5 file In-Reply-To: References: , <529F5CCA.5090200@iwr.uni-heidelberg.de> Message-ID: <52A02ADD.5070705@iwr.uni-heidelberg.de> Dear Bohumil, On 12/04/2013 07:22 PM, Maco Bohumil wrote: > I carve (segment) let say 20 object in one session. So I have 20 seeds with random colors. > Then I export the meshes to obj files and change the color of each seed to black. > Than I hit segmentation so I have all objects colorized black. > Then I export segmentation overlay into h5 file. > Next time I continue with the carving, > I import the segmentation from file (keeping color table from file) into ilastik. > And this is the reason to have all object from previous session to have the same color > in order to not interfere with seeds color of ongoing session (not mention that > I can place to all black are the background seeds straightaway). There are ways to fix the colortable, but it is going to require a bit of python scripting. I'll sketch a solution here, let me know if this was helpful or not: Let the exported overlay be "ov.h5". Then, in python: import h5py,numpy,pickle f = h5py.File("ov.h5", 'r') d = f["volume/data"] #the stored segmentation ctable = d.attrs["overlayColortable"] ctable = pickle.loads(ctable) Now you have the colortable "ctable" of the overlay. Looking at an overlay I just created with ilastik 0.5, it seems that the colortable is encoded in the following format: ctable[0] and ctable[1]: ignore these entries ctable[2] until ctable[N]: these are the colors of your carved objects The color is stored as ARGB values. So, for example hex(ctable[2]) = 0xff00ff00L #for a green object I carved The first "ff" means the layer is fully opaque, the "00" means no red, the "ff" means all of green and the "00" means no blue. Now you could alter this colortable to be all black (hopefully the value would be 0xff000000L), and write it back to the file via f = h5py.File("ov.h5", 'a') d.attrs["overlayTable"] = pickle.dumps(ctable) f.close() Now, loading this file should produce a different colortable... > And if I have so many of those h5 files this was the reason to find out the possibility > to merge them in a single one (e.g. one h5 file for all dendrites, one for all axons, > and so on). Next time I continue I load just one or two overlays... Again, combining all the carved objects would be easy using a small python scripts. Let's say you have all your segmentation .h5 files stored in a directory called "segs". Then (from the top of my head), I would do the following: import glob, h5py, numpy fnames = sorted(glob.glob("segs/*.h5")) #grab first segmentation to find out about the dimensionality of the #dataset f = h5py.File(fnames[0], 'r') sh = f["volume/data"].shape f.close() #create combined segmentation (white for all objects, otherwise black) combined = numpy.zeros(sh, dtype=numpy.uint8) for fname in fnames: f = h5py.File(fname, 'r') s = f["volume/data"].value combined[numpy.where(s>=2)] = 255 f.close() f = h5py.File("combined.seg", 'w') f.create_dataset("volume/data", data=combined) f.close() > > And as I said, sometimes I was in hurry and I simply just exported the segmentation overlay without changing the color of the seeds to the black. So I was seeking the option to change the color somehow editing the h5 file. > > Hope it make more sense now. Anyway I will explore the new version of the ilastik in the meantime. > > Again, thank you very much for your fast response. > > > PS: Friend of mine just e-mailed me about the new version of ilastik too and he mentioned that the down sampling volume option during loading the stack is missing. Is this true? Yes, this function is missing. I would suggest using fiji to create a downscaled variant of your image stack and then work on that. All the best, Thorben > > > ________________________________________ > From: Thorben Kroeger [thorben.kroeger at iwr.uni-heidelberg.de] > Sent: Wednesday, December 04, 2013 5:48 PM > To: Maco Bohumil > Cc: ilastik-user at ilastik.org; Thorben Kroeger > Subject: Re: [ilastik-user] segmentation overlay h5 file > > Hello Bohumil, > > On 12/04/2013 09:49 AM, Maco Bohumil wrote: >> it is possible to merge multiple h5 files of the segmentation overlay to >> the single one? > > You are working with ilastik version 0.5 and the carving workflow, correct? > >> I am working on a project where we are reconstructing all the objects in >> our large volume of FIBSEM image stacks. At the end of the session I >> save the overlay segmentation as h5 file in order to keep track of the >> object already segmented. > > So you are segmenting one object at a time? > There are two issues: > 1) the segmentation volume is not compressed, thus significant storage > space is needed > 2) objects may overlap (as they are extracted independently) > > In the new version of ilastik, 1.0 (download at > http://ilastik.github.io/index10.html) we have re-written carving to > support the workflow you are currently using. > > Each carved object can be saved under a name; it will be stored > compressed. ilastik can show all completed objects, they can be loaded > again for correctsions and so on. > Please have a look at > http://ilastik.github.io/documentation/carving/carving.html for the > documentation of carving in version 1.0 > > When I start new session in ilastik I import >> that segmentation overlay from h5 file. As the project is growing I have >> many h5 files and loading them one by one is not very convenient also >> ending up with long list to scroll. >> >> So I was wondering if is possible to merge those h5 files into a single >> one (all are from the same volume e.g. the same dimensions). > > The best way (other than switching to the new version) > would be to make a colored volume image. For each carved object, pick a > random color. Then write an HDF5 file > (dimensions (1,#x,#y,#z,3) with a RGB color for each x,y,z voxel. The > remaining problem is what to do when your objects overlap, but this is a > good check anyway in order to see how precise your carving is (if the > overlap is significant, the tracer likely made a mistake like missing a > branch). > >> >> And the second question: Sometimes I was in a hurry not to change the >> color of seeds to the same black color for all the seeds and during >> importing none of the option (use color from the file, use grey or use >> selected color (as black)) works. It is possible to change seeds color >> somehow editing the h5 file? > > I'm not sure I understand. If you forgot to change the seed color, and > then you load such a carved object back as .h5 file, the object appears > in another color? Why does it matter what color the object has? > > > All the best, > > Thorben > From kumartr at gmail.com Tue Jul 23 05:11:22 2013 From: kumartr at gmail.com (kumartr at gmail.com) Date: Mon, 22 Jul 2013 20:11:22 -0700 (PDT) Subject: [ilastik-user] Reg: Marking multiple (modality) images Message-ID: <11023244.0.1374549081977.JavaMail.javamailuser@localhost> An HTML attachment was scrubbed... URL: From kumartr at gmail.com Tue Jul 23 05:15:36 2013 From: kumartr at gmail.com (kumartr at gmail.com) Date: Mon, 22 Jul 2013 20:15:36 -0700 (PDT) Subject: [ilastik-user] Features in Ilastik Message-ID: <3555712.1.1374549335721.JavaMail.javamailuser@localhost> An HTML attachment was scrubbed... URL: From christoph.sommer at imba.oeaw.ac.at Tue Jul 23 09:25:13 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Tue, 23 Jul 2013 09:25:13 +0200 Subject: [ilastik-user] Reg: Marking multiple (modality) images In-Reply-To: <11023244.0.1374549081977.JavaMail.javamailuser@localhost> References: <11023244.0.1374549081977.JavaMail.javamailuser@localhost> Message-ID: <51EE2FD9.2090902@imba.oeaw.ac.at> Dear Kumar, thanks for you email! > Is there a way to copy markings of one image onto other 'modalities' > if required ? There is the possibility to 'inject' user annotation labels into the ilastik project (ilp) file. I attached an example python script. You need python, numpy, vigra, and h5py installed [1] > One option that I thought was to make a '3D' image using all the > layers and then load it into ialstik. I would not recommend this procedure, since ilastik pixel features are computed 3D (isotropically) and the data you are dealing has 'modalities' or spectra. > Would this be appropriate ? Are there alternative and efficient ways > to learn from multiple 'modality' or 'layers' ilastik can deal with multi-spectral data! Unfortunately, there is no GUI option to read multi-spectral data with more than 3 channels, but you can construct it per script (similar as in the attached script). Let me know if you need more information how to do that exactly. All the best, Chris [1] http://www.lfd.uci.edu/~gohlke/pythonlibs/ > > > Mail sent via way2sms. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- import vigra import h5py import numpy """ Example how to insert labels into an ilastik project files Dimensions in this example are x,y,z == 100,100,50 """ # open project file f = h5py.File('.ilp','a') # get file names of labels # you can also use the python module glob to read the file names from a folder label_file_names = [' References: <5295098.5.1374649590354.JavaMail.javamailuser@localhost> Message-ID: <51EF9496.9070405@imba.oeaw.ac.at> Dear Kumar, I had a look at your code. the GaussianBlur function of OpenCV expects "data, (kernelsizeX, kernelsizeY), sigma" as first three arguments. The unpacking with "applyToAllChannels" does not work for this case so I just overwrote this function with an appropriate one. I have attached a working code snipet to this email. Hope this works for you! Best, Christoph PS: your emails are rather difficult to decipher, since all comes in a single line :). What kind of client are you using? > Dear Chris Thanks for your prompt response I had a look at the > standardFeatures.py and also the Youtube video for 'Programming custom > features' > http://www.youtube.com/watch?feature=player_embedded&v=wD0a32F-Lhs I > am able to add features which are part of vigra (like Morphology > operations) I tried to use opencv2 functionalities, with a simple > example for GaussianBlur after importing opencv2 This does not seem to > compute any output, (snippet code below) I understand that we have to > provide a functor,as defined as 'func' 2. But what are the > requirements for such a 'func' is not clear What arguments must 'func' > return ? 3. If a 'func' does not have argument sigma or does not have > capability to be computed at different sigma , how is this handled > ------------------ import cv2 import cv2.cv as cv class > Sift(FeatureBase): name = "New" groups = ['New'] numOutputChannels2d = > 1 numOutputChannels3d = 1 def __init__(self, sigma): > FeatureBase.__init__(self,sigma) self.minContext = > int(numpy.ceil(sigma * 3.5)) def compute2d(self, data): func = > cv2.GaussianBlur result = self.applyToAllChannels(data, func, > self.sigma) return result def compute3d(self, data): func = > cv2.GaussianBlur result = self.applyToAllChannels(data, func, > self.sigma) return result -------------------------- Dear Kumar, After > you have cloned the ilastik source code from github, you can find the > file "FeatureBase.py" in > "ilasitk/modules/classification/core/features/". To create a new > feature, you simply have to inherit from the class "FeatureBase" and > implement the "compute" functions. For more details see the in-line > documentation. In the file "standardFeatures.py" you find examples of > all currently implemented ilastik pixel features. Cheers, Chris Warm > Regards Kumar > > > Mail sent via way2sms. -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at imba.oeaw.ac.at Fri Jun 7 10:04:26 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Fri, 7 Jun 2013 10:04:26 +0200 Subject: [ilastik-user] [Team] ilastik - CellProfiler bridge In-Reply-To: <51B1920C.2080708@nld.ds.mpg.de> References: <51AEF369.7010306@nld.ds.mpg.de> <51B08E19.6030209@imba.oeaw.ac.at> <51B1920C.2080708@nld.ds.mpg.de> Message-ID: <51B1940A.9050506@imba.oeaw.ac.at> Dear Lars, is using the CP Linux bundle also an option for you? If no, you have to install ilastik 0.5 from source into your PYTHONPATH. The most important dependency is vigra. git checkout: https://github.com/ilastik/ilastik-0.5 https://github.com/ilastik/vigra-ilastik-05 You can install vigra with CMake and ilastik via setuptools. @"linux-users of the user mailing list": I haven't done this for a while, so in case you have specific advice for Lars don't hesitate! :) Cheers, Christoph > Dear Christoph, > thanks for the quick response. I am using Linux and therefore had to run > CellProfiler from source. The CellProfiler version is 2.0.0 rev SVN: > 10386. > I am running Cellprofiler by calling the python script CellProfiler.py > which gives the error message 'No module named nuageux'. Despite this > CellProfiler is running and all existing modules seem to work. > Best, Lars From shahab.ensafi at nus.edu.sg Sat Jun 8 19:16:45 2013 From: shahab.ensafi at nus.edu.sg (Shahab Ensafi) Date: Sun, 9 Jun 2013 01:16:45 +0800 Subject: [ilastik-user] label image Message-ID: <78709A58-5256-40D2-AA42-0F22347B3B49@nus.edu.sg> Hi, I want to load my label image which is a tif file to ilastik to use in learning the classifier. do we have such capability or we should manually select the labels in the main image? Thank you for your help in advance. Best regards, From christoph.sommer at imba.oeaw.ac.at Mon Jun 10 10:39:41 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 10 Jun 2013 10:39:41 +0200 Subject: [ilastik-user] label image In-Reply-To: <78709A58-5256-40D2-AA42-0F22347B3B49@nus.edu.sg> References: <78709A58-5256-40D2-AA42-0F22347B3B49@nus.edu.sg> Message-ID: <51B590CD.4050401@imba.oeaw.ac.at> Dear Shahab Ensafi, the easiest way to insert labels from tif files into ilastik is to create an ilastik project file (.ilp) and use the attached script. When you create the .ilp file, also add the amount of label classes you would like to use in the Classification tab. It is important that the labels in your tif files are enumerated starting from zero. 1 == label 1, 2 == label 2, and so on. Zero means unlabeled. Hope this works for you!? Best regards, Chris > Hi, > I want to load my label image which is a tif file to ilastik to use in learning the classifier. > do we have such capability or we should manually select the labels in the main image? > > Thank you for your help in advance. > Best regards, > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- import vigra import h5py import numpy """ Example how to insert labels into an ilastik project files Dimensions in this example are x,y,z == 100,100,50 """ # open project file f = h5py.File('.ilp','a') # get file names of labels # you can also use the python module glob to read the file names from a folder label_file_names = [' Hi, I am new to ilastik and am trying to use it for classification of eye datasets I had a generic question, In the image there are several classes, like blood vessels, lesions etc I am mainly interested in the lesions Option 1: Do I then make a two class classifier, as lesion and non-lesion (and club all non-lesions region into this class), Option 2: Make a multi-class classifier, with blood vessel as one class for optic disk, and one for lesion, Which would give a better classification to detect lesions ? Kumar Rajamani Conversation with ilastik-user at ilastik.org, christoph.sommer at imba.oeaw.ac.at and Kumar Rajamani -------------- next part -------------- An HTML attachment was scrubbed... URL: From bernhard.kausler at iwr.uni-heidelberg.de Tue Jun 25 14:03:47 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Tue, 25 Jun 2013 14:03:47 +0200 Subject: [ilastik-user] Conversation with Kumar Rajamani and others In-Reply-To: <1372151504810:1319290013-3867880597@mail.projektitan.com> References: <1372151504810:1319290013-3867880597@mail.projektitan.com> Message-ID: <51C98723.7090400@iwr.uni-heidelberg.de> Dear Kumar, unfortunately there is no--in general--superior approach. You need to try both variants and see which performs better for your specific requirements. Personally, I would go for two classes since it reduces the combinatoric possibilities of unfavorable overlaps in feature space (that is, if you have - say - 100 different classes there would be 99 possibilities to choose the wrong class for a given pixel instead of 1 in the two classes case). Sorry, that I can't help you there effectivly. However, if you have more questions, feel free to ask :) Best, Bernhard On 25.06.2013 11:11, Kumar Rajamani wrote: > Hi, > I am new to ilastik and am trying to use it for classification of eye > datasets > > I had a generic question, > In the image there are several classes, like blood vessels, lesions etc > I am mainly interested in the lesions > Option 1: > Do I then make a two class classifier, as lesion and non-lesion (and > club all non-lesions region into this class), > Option 2: > Make a multi-class classifier, with blood vessel as one class for optic > disk, and one for lesion, > > Which would give a better classification to detect lesions ? > > Kumar Rajamani > > > Conversation with ilastik-user at ilastik.org > , christoph.sommer at imba.oeaw.ac.at > and Kumar Rajamani > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From ap at solarrain.com Sun Jun 30 00:47:20 2013 From: ap at solarrain.com (ap) Date: Sat, 29 Jun 2013 18:47:20 -0400 Subject: [ilastik-user] WIndows Behavior, 3D View, Zoom, etc. Message-ID: <003901ce751a$9df54720$d9dfd560$@solarrain.com> This is my feedback after a few hours with ilastik, great effort, great integration of python capabilities. Windows 7, 64 bit ilastik 0.5-win-v0.5.1 1. Stack of PNG's show up correctly in other views, but not in 3D view 2. CTRL mouse wheel only zooms one window not all three as in example videos. 3. It seems the ilp files are storing the data as 32 bit floating point though it was originally 8 bit unsigned. Low priority: 4. Store last input folder visited so user does not have to navigate each time they want to open a data set in the same or nearby folder. 5. Open H5 file types as an optional input to stack of images. -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image004.png Type: image/png Size: 2810 bytes Desc: not available URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image002.jpg Type: image/jpeg Size: 35053 bytes Desc: not available URL: From jason.maynes at sickkids.ca Wed Mar 27 20:44:41 2013 From: jason.maynes at sickkids.ca (Jason Maynes) Date: Wed, 27 Mar 2013 19:44:41 +0000 Subject: [ilastik-user] image types Message-ID: <0D15BC0F-B0EB-4F66-BA00-8F054585CED7@sickkids.ca> Hello: Is there a specific image type or bit that ilastik will or will not read? I seem to have trouble reading in images sometimes where the imaging pixel scaling appears to be off (i.e. the image shown by ilastik is either almost all black or all white, only showing the extreme peaks or valleys of the pixels intensities). This appears to be worse for png images and for images processed in some way by imagej first (i.e. if I read a tiff directly from my robotic imager, ilastik does fine, but if I read it into imagej, save it as a tiff again with a different filename, ilastik won't properly read it in). There is no error message displayed, as I said it just looks like the intensity scaling is all off. Could this be a problem with image headers put out by imageJ? Thanks. Cheers, JTM **************************************** "...everything that living things do can be understood in terms of the jigglings and wigglings of atoms." Richard Feynman, Nobel Laureate 1965 "We can be sure that if a detailed understanding of the molecular basis of chemo-therapeutic activity were to be obtained, the advance of medicine would be greatly accelerated." Linus Pauling, Nobel Laureate 1954 **************************************** Jason Thomas Maynes, PhD/MD Departments of Anaesthesia and Pain Medicine, Molecular Structure and Function and Biochemistry Hospital for Sick Children/SickKids Research Institute Faculty of Medicine University of Toronto Office: 416-813-5934 Cell: 647-746-6198 jason.maynes at sickkids.ca **************************************** ________________________________ This e-mail may contain confidential, personal and/or health information(information which may be subject to legal restrictions on use, retention and/or disclosure) for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies. -------------- next part -------------- An HTML attachment was scrubbed... URL: From bernhard.kausler at iwr.uni-heidelberg.de Thu Mar 28 10:15:26 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Thu, 28 Mar 2013 10:15:26 +0100 Subject: [ilastik-user] image types In-Reply-To: <0D15BC0F-B0EB-4F66-BA00-8F054585CED7@sickkids.ca> References: <0D15BC0F-B0EB-4F66-BA00-8F054585CED7@sickkids.ca> Message-ID: <51540A2E.3030708@iwr.uni-heidelberg.de> Dear Jason, my first guess would be that these images have a higher dynamic range than 8bit . Ilastik 0.5. only supports 8bit images (we fixed that in the upcoming release 0.6). Could you please send us one of the problematic images? We could help you much better when we can investigate the problem ourselves. If you don't want to post it in public, you can also send it directly to me: bernhard.kausler at iwr.uni-heidelberg.de Best, Bernhard On 03/27/2013 08:44 PM, Jason Maynes wrote: > Hello: > > Is there a specific image type or bit that ilastik will or will not read? > I seem to have trouble reading in images sometimes where the imaging > pixel scaling appears to be off (i.e. the image shown by ilastik is > either almost all black or all white, only showing the extreme peaks or > valleys of the pixels intensities). This appears to be worse for png > images and for images processed in some way by imagej first (i.e. if I > read a tiff directly from my robotic imager, ilastik does fine, but if I > read it into imagej, save it as a tiff again with a different filename, > ilastik won't properly read it in). There is no error message displayed, > as I said it just looks like the intensity scaling is all off. Could > this be a problem with image headers put out by imageJ? Thanks. > > Cheers, > JTM > > **************************************** > "...everything that living things do can be understood in terms of the > jigglings and wigglings of atoms." > > Richard Feynman, Nobel Laureate 1965 > > > "We can be sure that if a detailed understanding of the molecular basis of > chemo-therapeutic activity were to be obtained, the advance of medicine > would be greatly accelerated." > > Linus Pauling, Nobel Laureate 1954 > **************************************** > Jason Thomas Maynes, PhD/MD > Departments of Anaesthesia and Pain Medicine, Molecular Structure and > Function and Biochemistry > Hospital for Sick Children/SickKids Research Institute > Faculty of Medicine > University of Toronto > Office: 416-813-5934 > Cell: 647-746-6198 > jason.maynes at sickkids.ca > **************************************** > > > > > ------------------------------------------------------------------------ > > This e-mail may contain confidential, personal and/or health > information(information which may be subject to legal restrictions on > use, retention and/or disclosure) for the sole use of the intended > recipient. Any review or distribution by anyone other than the person > for whom it was originally intended is strictly prohibited. If you have > received this e-mail in error, please contact the sender and delete all > copies. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From ilastik-announce-bounces at ilastik.org Thu May 2 12:36:11 2013 From: ilastik-announce-bounces at ilastik.org (ilastik-announce-bounces at ilastik.org) Date: Thu, 02 May 2013 12:36:11 +0200 Subject: [ilastik-user] Forward of moderated message Message-ID: An embedded message was scrubbed... From: "Kirschmann, Moritz" Subject: Seeded Watershed for huge stacks Date: Tue, 30 Apr 2013 08:14:12 +0000 Size: 2391 URL: From christoph.straehle at iwr.uni-heidelberg.de Thu May 2 15:01:05 2013 From: christoph.straehle at iwr.uni-heidelberg.de (Christoph N. Straehle) Date: Thu, 02 May 2013 15:01:05 +0200 Subject: [ilastik-user] Forward of moderated message In-Reply-To: References: Message-ID: <51826391.3050802@iwr.uni-heidelberg.de> Hello Moritz, given your machine, volumes as large as yours might be usable. However, we usually divide our data sets into blocks of a size of 500x500x500 pixels. This ensures interactive response times. Greetings, Christoph On 05/02/2013 12:36 PM, ilastik-announce-bounces at ilastik.org wrote: > Dear ilastik team! > > I would like to do seeded watershed segmentation of neurons in relatively big stacks (2k x 2k x 2k and bigger). Is there a way to to it? > My computer with 128GB of RAM starts to chokes when I look at stacks of 2k x 2k x 400 voxel using only on feature (medium edges). > > Thank you very much for your effort! > > Best regards from Basel, Moritz From bernhard.kausler at iwr.uni-heidelberg.de Fri May 3 13:16:26 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Fri, 03 May 2013 13:16:26 +0200 Subject: [ilastik-user] Feasibility of Real-Time Video Stream Processing In-Reply-To: References: Message-ID: <51839C8A.30200@iwr.uni-heidelberg.de> Dear Darren, ilastik is as a GUI application to analyze large, multidimensional image data in an offline manner. It was not written with your intended application in mind. Unfortunately, we are also not familiar with real time image processing in a robotics context and can't give you any advice in that regard. Sorry, that I can't help you more. Your project sounds very exciting and I hope you can finish it in one way or another :) Best, Bernhard From ioannis.moutsatsos at novartis.com Fri May 10 15:49:01 2013 From: ioannis.moutsatsos at novartis.com (Moutsatsos, Ioannis) Date: Fri, 10 May 2013 13:49:01 +0000 Subject: [ilastik-user] Image appears overexposed Message-ID: Greetings; Thanks for a great and easy to use toolkit! Using InCell Analyzer 2000 TIFF images. When image is opened in the ilastic project view it appears overexposed while in other software it does not. This makes it difficult to paint the areas of interest for classification. Any suggestions of adjustments that can be made? For example [cid:image001.png at 01CE4D63.979E2D30] Best regards Ioannis K. Moutsatsos, Ph.D., M.S.E. Senior Research Investigator Novartis Institute for Biomedical Research Developmental & Molecular Pathways 250 Massachusetts Avenue - 4B-120-02 Cambridge, MA 02139 Tel 617.871.4269 ioannis.moutsatsos at novartis.com -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.png Type: image/png Size: 128568 bytes Desc: image001.png URL: From christoph.sommer at imba.oeaw.ac.at Mon May 13 10:38:47 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 13 May 2013 10:38:47 +0200 Subject: [ilastik-user] [Team] Fwd: Export classifier In-Reply-To: <518D3FE9.4090703@iwr.uni-heidelberg.de> References: <518D3FE9.4090703@iwr.uni-heidelberg.de> Message-ID: <5190A697.5020006@imba.oeaw.ac.at> Dear Carolina, the export classifier button is in the classification tab (upper right corner). If it is not, which version of ilastik are you using? All the best, Christoph > Hi! > I've been using Ilastik for tissue classification, and I'm very happy > with how it performs. > Now I'd like to use my classifier in CellProfiler, and I've been > looking at the instructions here: > http://ilastik.org/index.php?cat=20_Documentation&page=03_Cellprofiler > They seem very nice and clear, but the problem is I do not find a way > (button) to export the classifier. Can you please help? > Thanks > Carolina From christoph.sommer at imba.oeaw.ac.at Mon May 13 10:44:53 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 13 May 2013 10:44:53 +0200 Subject: [ilastik-user] [Team] Fwd: RE: Image appears overexposed In-Reply-To: <518D45D6.9090509@iwr.uni-heidelberg.de> References: <518D45D6.9090509@iwr.uni-heidelberg.de> Message-ID: <5190A805.1070303@imba.oeaw.ac.at> Dear Ioannis, InCell uses 12bits in a container 16bit tif. Can you try to use the "Normalization Overlay" (click the "plus" button below the overlay list). As minimum select 0 and es maximum 2^12 -1 = 4095. This should make the display right. Internally, ilastik handles the data correctly. Best, Christoph > Hello Ioannis, > > thanks for the encouraging comments... > > About your question: are your images 8 or 16 bit? Could you please > send us a test image (as an attached file)? > > Best regards, > > Fred From christoph.sommer at imba.oeaw.ac.at Thu May 23 17:43:38 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Thu, 23 May 2013 17:43:38 +0200 Subject: [ilastik-user] Fwd: problem starting on windows 7 In-Reply-To: <111C1E43DB683D46A1D382DE1909B39409821E72@MLBXv02.nih.gov> References: <111C1E43DB683D46A1D382DE1909B394068B5CA3@MLBXv02.nih.gov> <519B6D20.3080807@iwr.uni-heidelberg.de> <519B7290.7000806@imba.oeaw.ac.at> <111C1E43DB683D46A1D382DE1909B39409821E72@MLBXv02.nih.gov> Message-ID: <519E392A.4090401@imba.oeaw.ac.at> Dear Ghadi, thank you for your prompt answer! > I did install the software using the installer, but it did not result in an entry in the start menu. okay, there we got perhaps the problem! Apparently, the installer was not able to copy relevant files. Did you observe a error message during installation? Did you run the installer with administration privileges? If you don't have admin permissions you can also install ilastik 0.5 into user folder (e.g. below Documents or in your home directory) > If I'm in the ilastik directory, this > > C:\Program Files (x86)\ilastik-0.5.06-rc5>echo %PYTHONPATH% > C:\Program Files (x86)\ilastik-0.5.06-rc5 This is correct. > I also looked under Environment variables in System Properties, and I do not have a PYTHONPATH variable. All I have is PYTHONHOME variable set to C:\csvn\Python25 Sorry, I forgot to mention to echo the PYTHONPATH after you tried to execute run-ilastik.bat. It is set in the script. > C:\Program Files (x86)\ilastik-0.5.06-rc5>run-python This looks to me that you are running the 32bit version of ilastik. Are you sure that your Windows 7 is 32bit?? If it is 64bit, I _strongly_ recommend to use ilastik 0.5.12 for 53bit machines. http://www.ilastik.org/cms/download.php?cat=15_Download&file=ilastik-0.5-win-v0.5.12-x86-64.exe Hope this will work! Cheers, Chris >> >> Hi all, >> >> I installed Ilastik on my Windows 7 (64-bit) machine. I have the 2008 >> and 2010 redistribution packages in place. But, when I start a cmd >> window to run ilastik, I get the following: >> >> C:\Program Files\ilastik-0.5.12>run-ilastik >> >> Traceback (most recent call last): >> >> File "ilastik\ilastikMain.py", line 30, in >> >> import vigra >> >> ImportError: No module named vigra >> >> 'sleep' is not recognized as an internal or external command, >> >> operable program or batch file. >> >> When I check the path, only the dlls of ilastik shows up (see lines >> below). >> >> C:\Program Files\ilastik-0.5.12>path >> >> PATH=C:\Program Files\ilastik-0.5.12\dlls; [... etc] >> >> I've added the whole ilastik directory to the path, and that didn't help. >> >> This looks like a powerful software, I'm anxious to use it. Please help. >> >> Ghadi >> From hansmeetschool at gmail.com Fri Nov 15 01:29:18 2013 From: hansmeetschool at gmail.com (Hansmeet Singh) Date: Thu, 14 Nov 2013 16:29:18 -0800 Subject: [ilastik-user] Question on 3D reconstruction Message-ID: Hey guys, Right now I have a script to take in an ilatik probability, threshold it and clean it up a bit: ilastik_filename = img_filename.replace('.png', '.png_processed.h5') prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', 'r') label_index = 1 mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] prob_file.close() blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) mito_pred2 = blur_img<.85 mito_pred2 = mahotas.erode(mito_pred2, disc) but I'm having trouble (i.e. I don't know how to do this) with using this information, or code for a whole stack, to join up each prediction image and make a 3D reconstruction of the thresholded structures. This is a very open-ended question and, after not having success trying to find this information for a while I decided I'd just go ahead and mail the group. Thanks, Hansmeet -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at imba.oeaw.ac.at Fri Nov 15 10:05:59 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Fri, 15 Nov 2013 10:05:59 +0100 Subject: [ilastik-user] Question on 3D reconstruction In-Reply-To: References: Message-ID: <5285E3F7.1030808@imba.oeaw.ac.at> Dear Hansmeet, I am not sure if I understand you correctly, but if you want to clean up the entire z-stack of the prediction, you can use something like pred = prob_file['/volume/prediction'] for z in xrange(pred.shape[1]): zslice = pred[0,z,:,:,label_index] [... blur the zslice as you did before] You could also do the smoothing, threshold and erode directly in 3D. I dont know if skimage and mahotas allow 3D processing, but the "vigra" library certainly does. Cheers, Christoph > Hey guys, > > Right now I have a script to take in an ilatik probability, threshold > it and clean it up a bit: > > ilastik_filename = img_filename.replace('.png', '.png_processed.h5') > prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', > 'r') > label_index = 1 > mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] > prob_file.close() > blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) > mito_pred2 = blur_img<.85 > mito_pred2 = mahotas.erode(mito_pred2, disc) > > but I'm having trouble (i.e. I don't know how to do this) with using > this information, or code for a whole stack, to join up each > prediction image and make a 3D reconstruction of the thresholded > structures. This is a very open-ended question and, after not having > success trying to find this information for a while I decided I'd just > go ahead and mail the group. > > Thanks, > Hansmeet > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From luca.fiaschi at iwr.uni-heidelberg.de Fri Nov 15 10:27:48 2013 From: luca.fiaschi at iwr.uni-heidelberg.de (Luca Fiaschi) Date: Fri, 15 Nov 2013 10:27:48 +0100 Subject: [ilastik-user] Question on 3D reconstruction In-Reply-To: References: Message-ID: <5285E914.6070503@iwr.uni-heidelberg.de> Hi, If I understand correctly: you would like to make a 3D volumetric rendering of the segmented (and eroded) probability map for the entire stack. This step requires to transform the segmentation volume into a mesh and then display the mesh. A very good tool to do something like this is Vaa3D http://www.vaa3d.org/. (You may need to export the segmented stack as a series of tiff files.) Otherwise depending it is not too difficult to program something in python. Good libraries for 3D rendering are Mayavi2 ( small volumes, but simple to use) or PyVTK (bigger volumes but a quite a bit more complex). Best Luca On 15.11.2013 01:29, Hansmeet Singh wrote: > Hey guys, > > Right now I have a script to take in an ilatik probability, threshold > it and clean it up a bit: > > ilastik_filename = img_filename.replace('.png', '.png_processed.h5') > prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', > 'r') > label_index = 1 > mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] > prob_file.close() > blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) > mito_pred2 = blur_img<.85 > mito_pred2 = mahotas.erode(mito_pred2, disc) > > but I'm having trouble (i.e. I don't know how to do this) with using > this information, or code for a whole stack, to join up each > prediction image and make a 3D reconstruction of the thresholded > structures. This is a very open-ended question and, after not having > success trying to find this information for a while I decided I'd just > go ahead and mail the group. > > Thanks, > Hansmeet > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -- Luca Fiaschi HCI/IWR - Heidelberg University Speyerer str. 6 - office H2.12 69115 Heidelberg Tel: +49 (6221) 54 - 5280 Fax: +49 (6221) 54 - 5276 -------------- next part -------------- An HTML attachment was scrubbed... URL: From hansmeetschool at gmail.com Fri Nov 15 19:42:29 2013 From: hansmeetschool at gmail.com (Hansmeet Singh) Date: Fri, 15 Nov 2013 10:42:29 -0800 Subject: [ilastik-user] Question on 3D reconstruction In-Reply-To: <5285E914.6070503@iwr.uni-heidelberg.de> References: <5285E914.6070503@iwr.uni-heidelberg.de> Message-ID: These options sound great; I'll look into them. Thanks for getting back to me on this very open-ended inquiry. - Hansmeet On Fri, Nov 15, 2013 at 1:27 AM, Luca Fiaschi < luca.fiaschi at iwr.uni-heidelberg.de> wrote: > Hi, > If I understand correctly: you would like to make a 3D volumetric > rendering of the segmented (and eroded) probability map for the entire > stack. This step requires to transform the segmentation volume into a mesh > and then display the mesh. > > A very good tool to do something like this is Vaa3D http://www.vaa3d.org/. > (You may need to export the segmented stack as a series of tiff files.) > > Otherwise depending it is not too difficult to program something in > python. > Good libraries for 3D rendering are Mayavi2 ( small volumes, but simple to > use) or PyVTK (bigger volumes but a quite a bit more complex). > > Best > Luca > > > > On 15.11.2013 01:29, Hansmeet Singh wrote: > > Hey guys, > > Right now I have a script to take in an ilatik probability, threshold it > and clean it up a bit: > > ilastik_filename = img_filename.replace('.png', '.png_processed.h5') > prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', > 'r') > label_index = 1 > mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] > prob_file.close() > blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) > mito_pred2 = blur_img<.85 > mito_pred2 = mahotas.erode(mito_pred2, disc) > > but I'm having trouble (i.e. I don't know how to do this) with using > this information, or code for a whole stack, to join up each prediction > image and make a 3D reconstruction of the thresholded structures. This is a > very open-ended question and, after not having success trying to find this > information for a while I decided I'd just go ahead and mail the group. > > Thanks, > Hansmeet > > > _______________________________________________ > ilastik-user mailing listilastik-user at ilastik.orghttp://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > > -- > Luca Fiaschi > HCI/IWR - Heidelberg University > Speyerer str. 6 - office H2.12 > 69115 Heidelberg > Tel: +49 (6221) 54 - 5280 > Fax: +49 (6221) 54 - 5276 > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From hansmeetschool at gmail.com Sat Nov 16 04:30:12 2013 From: hansmeetschool at gmail.com (Hansmeet Singh) Date: Fri, 15 Nov 2013 19:30:12 -0800 Subject: [ilastik-user] Question on 3D reconstruction In-Reply-To: References: <5285E914.6070503@iwr.uni-heidelberg.de> Message-ID: I'm looking into Vaa3D, and, I have to ask, when you say: You may need to export the segmented stack as a series of tiff files, do you mean that I should somehow turn mito_pred2 in this bit of code: ilastik_filename = img_filename.replace('.png', '.png_processed.h5') prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', 'r') label_index = 1 mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] prob_file.close() blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) mito_pred2 = blur_img<.85 mito_pred2 = mahotas.erode(mito_pred2, disc) into a tiff file (and then turn all of the segmented images - mito_pred3, mito_pred4 etc) into tiff files using python, after which I am to use Vaa3D to make the renderings? If so, do you have any suggestions as to what I am to code to accomplish this (ie how to turn mito_pred2 into a TIFF file). I'm sort of new to python programming so this kind of guidance really helps. Thanks, Hansmeet On Fri, Nov 15, 2013 at 10:42 AM, Hansmeet Singh wrote: > These options sound great; I'll look into them. Thanks for getting back to > me on this very open-ended inquiry. - Hansmeet > > > On Fri, Nov 15, 2013 at 1:27 AM, Luca Fiaschi < > luca.fiaschi at iwr.uni-heidelberg.de> wrote: > >> Hi, >> If I understand correctly: you would like to make a 3D volumetric >> rendering of the segmented (and eroded) probability map for the entire >> stack. This step requires to transform the segmentation volume into a mesh >> and then display the mesh. >> >> A very good tool to do something like this is Vaa3D http://www.vaa3d.org/ >> . >> (You may need to export the segmented stack as a series of tiff files.) >> >> Otherwise depending it is not too difficult to program something in >> python. >> Good libraries for 3D rendering are Mayavi2 ( small volumes, but simple >> to use) or PyVTK (bigger volumes but a quite a bit more complex). >> >> Best >> Luca >> >> >> >> On 15.11.2013 01:29, Hansmeet Singh wrote: >> >> Hey guys, >> >> Right now I have a script to take in an ilatik probability, threshold >> it and clean it up a bit: >> >> ilastik_filename = img_filename.replace('.png', '.png_processed.h5') >> prob_file = h5py.File('Thousands_mito_em_s1152.png_processed (1).h5', >> 'r') >> label_index = 1 >> mito_prob = prob_file['/volume/prediction'][0,0,:,:,label_index] >> prob_file.close() >> blur_img = scipy.ndimage.gaussian_filter(mito_prob, 13) >> mito_pred2 = blur_img<.85 >> mito_pred2 = mahotas.erode(mito_pred2, disc) >> >> but I'm having trouble (i.e. I don't know how to do this) with using >> this information, or code for a whole stack, to join up each prediction >> image and make a 3D reconstruction of the thresholded structures. This is a >> very open-ended question and, after not having success trying to find this >> information for a while I decided I'd just go ahead and mail the group. >> >> Thanks, >> Hansmeet >> >> >> _______________________________________________ >> ilastik-user mailing listilastik-user at ilastik.orghttp://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >> >> >> -- >> Luca Fiaschi >> HCI/IWR - Heidelberg University >> Speyerer str. 6 - office H2.12 >> 69115 Heidelberg >> Tel: +49 (6221) 54 - 5280 >> Fax: +49 (6221) 54 - 5276 >> >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: From neuhausmj at gmail.com Tue Nov 26 16:26:16 2013 From: neuhausmj at gmail.com (Miranda Segura) Date: Tue, 26 Nov 2013 09:26:16 -0600 Subject: [ilastik-user] Uploading 3D files other than images and stacks Message-ID: Hello, I am using a Nanofocus confocal 3D measurement system to generate thousands of leaf surface topographies. In addition to the topography file, the software also outputs a reflective intensity file (what our eyes see). From these measurements, I want to categorize two cell types, so Ilastik seems like it could do the trick- if I could upload the files. The files that are generated from this software include an .stl or .set file format. They can also generate .mnt or .nms files, but those are specific to Nanofocus software (I believe). Is there anyway I can upload and use other file types in Ilastik? Thanks for any help! -- Miranda Segura PhD Candidate Dept. of Plant Biology University of Illinois-Champaign Urbana -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Wed Nov 27 14:33:19 2013 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Wed, 27 Nov 2013 14:33:19 +0100 Subject: [ilastik-user] a question regarding Ilastik In-Reply-To: References: Message-ID: <5295F49F.8060100@iwr.uni-heidelberg.de> Dear Daphna, I'm glad to hear you are finding ilastik so useful for you. Unfortunately, we don't have a way to do volumetric measurements. To export the resulting segmentation and measure the objects in another program, try to use the "write segmentation" option in the automated classification tab. This will save the segmentation images rather than probability maps. There are also ways to save individual objects from seeded watershed (carving), but other people on the team are more familiar with them than I am. I've cc-ed my reply to our mailing list and I'll ask them to answer. Best, Anna On 27.11.2013 13:39, dafi sourani wrote: > Dea Dr. Kreshuk, > My name is Daphna, a PhD student for medical science using MRI. > I started using Ilastik with my MRI data about two weeks ago and I > found it very useful, friendly and helpful. > First, thank you for developng such a great program. > If possible, I would like to ask a question: My goal is to segment a > few importants organs and to calculate their volumes. > For the segmentation I used both automate classification and seeded > watershed (for different purposes). However, I wonder if there is a > tool which calculates the total volume of the segmented organs. > If not- how can I export the coordinates of the pixels containing the > volume in order to calculate the volume in Matlab or other programming > tool? > Thank you very much in advence, > All the best, > Daphna From christoph.sommer at imba.oeaw.ac.at Tue Sep 10 10:40:38 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Tue, 10 Sep 2013 10:40:38 +0200 Subject: [ilastik-user] [Team] question: how to apply a h5 classifier to a set of images In-Reply-To: <58D4FE65-BCAC-4736-A4B0-E80B68A2DB93@dzne.de> References: <520B7BAF.4050003@imba.oeaw.ac.at> <0645E7F4-D3DE-41E8-A70D-198BD4A002B8@dzne.de> <521B0727.5080105@imba.oeaw.ac.at> <58D4FE65-BCAC-4736-A4B0-E80B68A2DB93@dzne.de> Message-ID: <522EDB06.6020906@imba.oeaw.ac.at> Dear Christoph, in the CellProfiler module (classifyPixels) you can select the output class of ilastik. You have to specify the label index number after you selected the classifier file, which is just the enumerated number of the classes you have specified in ilastik. Best, Christoph > Hallo Ilastik Team, > > eine Frage zum h5 classifier import in cell profiler: > > Cell Profiler produziert eine probability map mit Hilfe des > classifiers. Doch wie w?hle ich die gew?nscht Klasse aus, wenn mein > Classifier auf mehr als 2 Klassen trainiert wurde? > > Viele Gr??e > Christoph > > > > > On Aug 26, 2013, at 9:43 AM, Christoph Sommer > > wrote: > >> Dear Christoph, >> >> concerning your question, when to export an h5 classifier file: >> >> You typically export the h5 classifier file if you want to: >> >> * use it together with CellProfiler [1] >> * use it together with CellCognition [2] >> * or if you want to integrate it into your self programmed pipeline >> >> There are also a few commercial software tools, which use the ilastik >> classifier file. >> >> Cheers, >> Christoph >> >> [1] http://www.cellprofiler.org/ >> [2] http://www.cellcognition.org/ >> >>> Hallo Christoph, >>> danke fuer deine Antwort! >>> Ich war im Urlaub, deswegen die sp?te Reaktion. >>> Eine Frage habe ich noch dazu: Wann macht es Sinn, den >>> Klassifizierer als h5-Datei zu exportieren? Wozu kann ich diese >>> Datei verwenden? >>> >>> Viele Gr??e >>> Christoph >>> >>> On Aug 14, 2013, at 2:44 PM, Christoph Sommer >>> >> > wrote: >>> >>>> Dear Christoph, >>>> >>>> the easiest way to apply a trained ilastik classification to more >>>> images is to use the "Automate Classification" tab. There, you can >>>> select in "Batch process" more images to be predicted (together >>>> with output-options). >>>> >>>> There is also an command line option for batch process. Let me >>>> know, if you need more information on that! >>>> >>>> Cheers, >>>> Christoph >>>> PS: sorry for the late reply!! >>>> >>>>> Dear Ilastik people, >>>>> >>>>> I'm currently using Ilastik in a cell segmentation project, where >>>>> Meera (the domain expert) trains a h5 classifier in ilastik. I aim >>>>> to apply this classifier on large datasets and produce probability >>>>> maps. >>>>> >>>>> How can I import a h5 classifier file to ilastik and run a batch >>>>> classification, outputting probability images, in ilastik? >>>>> >>>>> >>>>> The project is quite interesting, because it is a combination of >>>>> imageJ scripting (image conversion and filtering), ilastik >>>>> classification and definiens scripting (image segmentation and >>>>> calculation of object features) >>>>> >>>>> >>>>> Best >>>>> Christoph >>>>> >>>>> >>>>> >>>>> Dr. Christoph M?hl >>>>> Image and Data Analysis Facility Manager >>>>> Deutsches Zentrum f?r Neurodegenerative Erkrankungen e. V. (DZNE) >>>>> Ludwig-Erhard-Allee 2 >>>>> D- 53175 Bonn >>>>> >>>>> phone +49 228 43302 638 >>>>> mail christoph.moehl at dzne.de >>>>> >>>>> room C.OG.2.3.12 >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> team mailing list >>>>> team at ilastik.org >>>>> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/team >>>> >>> >>> Dr. Christoph M?hl >>> Image and Data Analysis Facility Manager >>> Deutsches Zentrum f?r Neurodegenerative Erkrankungen e. V. (DZNE) >>> Ludwig-Erhard-Allee 2 >>> D- 53175 Bonn >>> >>> phone +49 228 43302 638 >>> mail christoph.moehl at dzne.de >>> >>> room C.OG.2.3.12 >>> >>> >>> >>> >> > > Dr. Christoph M?hl > Image and Data Analysis Facility Manager > Deutsches Zentrum f?r Neurodegenerative Erkrankungen e. V. (DZNE) > Ludwig-Erhard-Allee 2 > D- 53175 Bonn > > phone +49 228 43302 638 > mail christoph.moehl at dzne.de > > room C.OG.2.3.12 > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From kumartr at gmail.com Wed Sep 11 17:59:55 2013 From: kumartr at gmail.com (Kumar Rajamani) Date: Wed, 11 Sep 2013 21:29:55 +0530 Subject: [ilastik-user] Reg: Import of Classifier Message-ID: Dear All, There exist option to save a classifier once it has been trained on some samples This is an export classifier which exists Unfortunately there is no feature to import a learnt classifier (from a earlier training using ilastik) This is generally needed if one happens to train/test over several days and wants to export the classifier This might be a good feature to be incorporated into ilastik in future realses Warm Regards Kumar -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at imba.oeaw.ac.at Thu Sep 12 10:53:32 2013 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Thu, 12 Sep 2013 10:53:32 +0200 Subject: [ilastik-user] Reg: Import of Classifier In-Reply-To: References: Message-ID: <5231810C.6040706@imba.oeaw.ac.at> Dear Kumar, to achieve this, simply save an ilastik project. Also in there the classifier is saved. The export function is just needed to interface to other tools such as CellProfiler or commercial software. Best, Christoph > learnt classifier (from a earlier training using ilastik) > This is generally needed if one happens to train/test over several > days and wants to export the classifier > This might be a good feature to be incorporated into ilastik in future > realses From corrado.cali at kaust.edu.sa Thu Sep 12 11:12:38 2013 From: corrado.cali at kaust.edu.sa (Corrado Cali) Date: Thu, 12 Sep 2013 09:12:38 +0000 Subject: [ilastik-user] Installing problems on MacBook Pro Message-ID: <783EBD0B-29EE-4616-BE5F-8077152CDF56@kaust.edu.sa> Dear all, I am contacting you because I am experiencing some problems during installation of Ilastik. I have been using your software on daily basis for almost 6 months now, and It always worked at its best in all the machines I have installed it (exclusively Mac for what I am concerned). Now I got a new machine, it's a macbook pro latest generation, 2.7 GHz Intel Core i7, 16GB ram, the OSX version is the 10.8.4. When I try to run the image, I have an error (please refer to the screenshot at the bottom) and the system won't open the image. Does anyone experienced the same problem before? Thank you in advance for any advice. Best regards, CC :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: Ing. Corrado Cal?, PhD Post Doctoral Fellow Laboratory of Cell Imaging and Energetics Division of BESE Room 4266, Level 4, Building 2 4700 King Abdullah University of Science and Technology Thuwal 23955-6900, Kingdom of Saudi Arabia Tel: +966 (0)2 808 2969 skype: menestrello82 Please note that week ends on the kingdom are on fridays and saturdays :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: [cid:EF7BB07E-EEBD-4F23-8A4F-8BA581EC83EB at KAUST.EDU.SA] ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: Screen Shot 2013-09-12 at 12.08.38 PM.png Type: image/png Size: 20870 bytes Desc: Screen Shot 2013-09-12 at 12.08.38 PM.png URL: From bernhard.kausler at iwr.uni-heidelberg.de Fri Sep 13 11:24:42 2013 From: bernhard.kausler at iwr.uni-heidelberg.de (Bernhard X. Kausler) Date: Fri, 13 Sep 2013 11:24:42 +0200 Subject: [ilastik-user] Installing problems on MacBook Pro In-Reply-To: <783EBD0B-29EE-4616-BE5F-8077152CDF56@kaust.edu.sa> References: <783EBD0B-29EE-4616-BE5F-8077152CDF56@kaust.edu.sa> Message-ID: <5232D9DA.60005@iwr.uni-heidelberg.de> Dear CC, we tested it on Mountain Lion (10.8) and didn't encounter any problems like yours. The error "disk image not recognized" could be caused by some corruption in your image file. Try to download the image again from our website. That could already fix the problem. Best, Bernhard On 09/12/2013 11:12 AM, Corrado Cali wrote: > Dear all, > > I am contacting you because I am experiencing some problems during > installation of Ilastik. > > I have been using your software on daily basis for almost 6 months now, > and It always worked at its best in all the machines I have installed it > (exclusively Mac for what I am concerned). > > Now I got a new machine, it's a macbook pro latest generation, 2.7 GHz > Intel Core i7, 16GB ram, the OSX version is the 10.8.4. When I try to > run the image, I have an error (please refer to the screenshot at the > bottom) and the system won't open the image. > > Does anyone experienced the same problem before? > > Thank you in advance for any advice. > > Best regards, > > CC > > > > :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: > /Ing.* **Corrado Cal?**, *PhD > Post Doctoral Fellow/ > *L*aboratory of *C*ell *I*maging and *E*nergetics > Division of BESE > Room 4266, Level 4, Building 2 > 4700 King Abdullah University of Science and Technology > Thuwal 23955-6900, Kingdom of Saudi Arabia > Tel:+966 (0)2 808 2969 > skype:menestrello82 > /Please note that week ends on the kingdom are on fridays and saturdays/ > :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: > > ------------------------------------------------------------------------ > > This message and its contents including attachments are intended solely > for the original recipient. If you are not the intended recipient or > have received this message in error, please notify me immediately and > delete this message from your computer system. Any unauthorized use or > distribution is prohibited. Please consider the environment before > printing this email. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From kumartr at gmail.com Sat Sep 14 10:18:12 2013 From: kumartr at gmail.com (Kumar Rajamani) Date: Sat, 14 Sep 2013 13:48:12 +0530 Subject: [ilastik-user] Reg: Import of Classifier Message-ID: Dear Christoph Thanks for your prompt response and for sharing your suggestions I have tried the option of saving the ilastik project and reloading it on restart As I observed, when restarted, the training information done in earlier session is lost and ilastik has to be retrained, This takes up time and is repeated effort Another observation is that, just like export helps in connecting bridging two applications (ilastik and CellProfiler) Similarly Import would help to bridge two users of ilastik One expert builds the classifier and saves as HF5 file using export Another regular user could use this by importing the HF5 classifier file and testing on other datasets Hence I suggested having a import of classifier could be a good option to have Also it would help to connect and use another Classifier like those within Fiji (ImageJ) and building over it using ilastik viz - (Fiji, ilastik connect) Warm Regards Kumar On Thu, Sep 12, 2013 at 2:23 PM, Christoph Sommer < christoph.sommer at imba.oeaw.ac.at> wrote: > Dear Kumar, > > to achieve this, simply save an ilastik project. Also in there the > classifier is saved. The export function is just needed to interface to > other tools such as CellProfiler or commercial software. > > Best, > Christoph > > > learnt classifier (from a earlier training using ilastik) >> This is generally needed if one happens to train/test over several days >> and wants to export the classifier >> This might be a good feature to be incorporated into ilastik in future >> realses >> > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From rniwayamembl1 at gmail.com Mon Sep 30 20:30:27 2013 From: rniwayamembl1 at gmail.com (Ritsuya Niwayama) Date: Mon, 30 Sep 2013 20:30:27 +0200 Subject: [ilastik-user] 3D batch Message-ID: Hello, ilastik team, This is Ritsuya, an end user. Can we perform 3d batch process using ilastik 0.5? I see that 2D batch is introduced in the documentation but 3D not. Furthermore as fur as I briefly tested, I could not. Sincerely yours, -------------- next part -------------- An HTML attachment was scrubbed... URL: From dlogan at broadinstitute.org Thu Apr 3 20:32:12 2014 From: dlogan at broadinstitute.org (David Logan) Date: Thu, 3 Apr 2014 14:32:12 -0400 Subject: [ilastik-user] Variable Importance Message-ID: Hi, I am on the CellProfiler team and am using ilastik 0.5 with CellProfiler. I found the getImportantVariables.py script and am using it for feature selection before I run lots of images in CellProfiler. The output barplot seems mostly reasonable (and helpful!), however at the "gigahuge" scale all feature categories show quite high importance, even though I don't expect some of them, e.g. orientation, to make sense. Two questions: (1) Should I trust this fairly old script? (2) Is there an updated script or procedure? I found this github issue ( https://github.com/ilastik/ilastik/issues/152) which implies there is not, though. Thanks for any help. Best, David -- David Logan, Ph.D. Computational Biologist Imaging Platform Broad Institute of MIT and Harvard 617-714-7755 www.broadinstitute.org/~dlogan -------------- next part -------------- An HTML attachment was scrubbed... URL: From hughesadam87 at gmail.com Fri Apr 4 21:40:00 2014 From: hughesadam87 at gmail.com (Adam Hughes) Date: Fri, 4 Apr 2014 15:40:00 -0400 Subject: [ilastik-user] Shape descriptors via vigra/ilastik Message-ID: Hello, I was reading through Feature Accumulators in VIGRA: ( http://ukoethe.github.io/vigra/doc/vigra/group__FeatureAccumulators.html#histogram ) It looks like the majority of these accumulators are based on pixel intensity information. I'm working with test images that intentionally have objects with randomly distributed brightnesses, but different shapes. Therefore, a descriptor such as "circularity" is really apt for distinguishing these objects in the binary image. Size alone is not quite enough to separate all of the species in the image. In some image processing libraries, shape descriptors are usually accessible through the API; however, I'm not quite clear on how or if some of the shape descriptors are implemented in Ilastik. Right now we have a classifier that uses Count (which is basically equivalent to pixel area if I understand?) and region-radii, which I assume is related to a circular fit of the area. Is there a feature that might be more related purely to shape like eccentricity that is accessible through the interface? Thanks -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Sun Apr 6 15:32:37 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sun, 06 Apr 2014 15:32:37 +0200 Subject: [ilastik-user] Variable Importance In-Reply-To: References: Message-ID: <20140406153237.fo6lsrwb28kggksc@wwwmail.urz.uni-heidelberg.de> Hi David, As long as you are using ilastik 0.5, the script should not be outdated. As you have seen in the github issue, updating it is on our todo list, but it's not there yet, sorry. I'll take a closer look at the script tomorrow to see if there is no bug which benefits the "gigahuge" features. Cheers, Anna Quoting David Logan : > Hi, > > I am on the CellProfiler team and am using ilastik 0.5 with CellProfiler. > I found the getImportantVariables.py script and am using it for feature > selection before I run lots of images in CellProfiler. The output barplot > seems mostly reasonable (and helpful!), however at the "gigahuge" scale all > feature categories show quite high importance, even though I don't expect > some of them, e.g. orientation, to make sense. > > Two questions: > (1) Should I trust this fairly old script? > (2) Is there an updated script or procedure? I found this github issue ( > https://github.com/ilastik/ilastik/issues/152) which implies there is not, > though. > > Thanks for any help. > Best, > David > -- > David Logan, Ph.D. > Computational Biologist > Imaging Platform > Broad Institute of MIT and Harvard > 617-714-7755 > www.broadinstitute.org/~dlogan > From christoph.sommer at imba.oeaw.ac.at Mon Apr 7 17:11:03 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 7 Apr 2014 17:11:03 +0200 Subject: [ilastik-user] Variable Importance In-Reply-To: References: Message-ID: <5342C007.4020806@imba.oeaw.ac.at> Hi David, > (1) Should I trust this fairly old script? Well, as far as I know, the way the variable importance is computed inside vigra has not changed, even though the script is rather old. @Anna: thanks for double checking!! > however at the "gigahuge" scale all feature categories show quite high > importance, even though I don't expect some of them, For me this sound like a fundamental issue. Gigahuge features take a (giga) huge neighborhood into account and one -in principle - needs more examples to learn such structures (just because of combinatorials)... Perhaps this is a side effect off over-fitting!? How does the classifier generalize on other images? Cheers, Christoph > Hi, > > I am on the CellProfiler team and am using ilastik 0.5 with > CellProfiler. I found the getImportantVariables.py script and am > using it for feature selection before I run lots of images in > CellProfiler. The output barplot seems mostly reasonable (and > helpful!), however at the "gigahuge" scale all feature categories show > quite high importance, even though I don't expect some of them, e.g. > orientation, to make sense. > > Two questions: > (1) Should I trust this fairly old script? > (2) Is there an updated script or procedure? I found this github > issue (https://github.com/ilastik/ilastik/issues/152) which implies > there is not, though. > > Thanks for any help. > Best, > David > -- > David Logan, Ph.D. > Computational Biologist > Imaging Platform > Broad Institute of MIT and Harvard > 617-714-7755 > www.broadinstitute.org/~dlogan > > > > > > > > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From ullrich.koethe at iwr.uni-heidelberg.de Mon Apr 7 19:05:06 2014 From: ullrich.koethe at iwr.uni-heidelberg.de (Ullrich Koethe) Date: Mon, 07 Apr 2014 19:05:06 +0200 Subject: [ilastik-user] Shape descriptors via vigra/ilastik In-Reply-To: References: Message-ID: <5342DAC2.2070403@iwr.uni-heidelberg.de> Hi Adam, Am 04.04.2014 21:40, schrieb Adam Hughes: > I was reading through Feature Accumulators in VIGRA: > > (http://ukoethe.github.io/vigra/doc/vigra/group__FeatureAccumulators.html#histogram) > > It looks like the majority of these accumulators are based on pixel > intensity information. VIGRA's accumulator framework implements a "mini language" for feature computation. By default, features apply to the intensities, but you can add modifiers to feature names in order to change their behavior. In your case, the modifier of interest is "Coord" which computes features on the basis of object shape. For example: "Mean" is the average intensity "Coord" is the center of the shape "RegionCenter" is an alias for "Coord". Similarly, "Coord" and "Coord" are the lower and upper bounds of the object's bounding box. One could also write "Coord", but this will obviously give the same result as plain "Count". The feature "Coord" gives the extend of the object shape along the coordinate axes. However, this is not an intrinsic object property, because the outcome depends on the relative orientation between the object and the axes. Usually, one is more interested in object size relative to the object's intrinsic coordinate system, i.e. aligned with its principal directions. Therefore, one usually writes "Coord>" to compute this feature in the principal coordinate system. "RegionRadii" is an abbreviation of this name. The eccentricity is commonly defined as sqrt(1 - b^2/a^2), where a and b denote maximum and minimum region radius respectively. Higher order properties are computed by "Coord>" and "Coord>" Skewness measures if the object is shaped asymmetrically, e.g. a pear would have higher skewness along its major axis than an apple. Kurtosis measures if the shape concentrates about its center -- a doughnut has higher kurtosis than a disc. You can also combine shape and intensity by computing weighted statistics where intensity is interpreted as mass. This is achieved by the "Weighted" modifier: "Weighted>" is the center of mass (alias: "CenterOfMass") "Weighted>" are the moments of inertia (alias: "MomentsOfInertia") "Weighted>" and "Weighted>" generalize the above features Two new features are available for weighted coordinates: "Coord" and "Coord" report the coordinate where the weight (i.e. intensity) assumes its minimum and maximum values respectively. Note that the accumulator framework computes all features by just scanning over the image one or more times (hence the name "accumulator"). Therefore, it only supports features that can be computed in this manner. This is a deliberate design descision to maximize speed and to allow blockwise feature computation. There are certainly many worthwhile shape features which cannot be computed in an accumulator fashion, but we found the shape features described above to perform quite well in various tasks. Hope this helps. Ullrich Koethe From bohumil.maco at epfl.ch Wed Apr 9 13:54:35 2014 From: bohumil.maco at epfl.ch (Maco Bohumil) Date: Wed, 9 Apr 2014 11:54:35 +0000 Subject: [ilastik-user] export meshes button in ilastik 1.0.3 Message-ID: Hi developers, I am testing the latest version of ilastik 1.0.3 for our reconstruction projects from FIBSEM volumes. But I have realized that the carving module for whatever reason does not have the option to save (export) meshes as obj file like the 0.5 version did. Please, could you let me know how I can get the reconstructed models out of ilastik 1.0.3? With a great improvement in the workflow of the recent version it is bit a shame not to be able get these models... Please, I really need this option quite urgent. Thank you very much for your help. Cheers, Bohumil -------------- next part -------------- An HTML attachment was scrubbed... URL: From bergs at janelia.hhmi.org Thu Apr 10 05:27:33 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Thu, 10 Apr 2014 03:27:33 +0000 Subject: [ilastik-user] export meshes button in ilastik 1.0.3 Message-ID: An HTML attachment was scrubbed... URL: From ullrich.koethe at iwr.uni-heidelberg.de Thu Apr 10 12:15:34 2014 From: ullrich.koethe at iwr.uni-heidelberg.de (Ullrich Koethe) Date: Thu, 10 Apr 2014 12:15:34 +0200 Subject: [ilastik-user] export meshes button in ilastik 1.0.3 In-Reply-To: References: Message-ID: <53466F46.1040302@iwr.uni-heidelberg.de> Hi Stuart, I created and uploaded Windows binaries for ilastik 1.0.5. I also updated the download links in ilastik.github.io/download.html. Best Ulli From dlogan at broadinstitute.org Fri Apr 11 19:13:34 2014 From: dlogan at broadinstitute.org (David Logan) Date: Fri, 11 Apr 2014 13:13:34 -0400 Subject: [ilastik-user] Variable Importance Message-ID: Thanks for the replies, Christoph. I will assume the getVariableImportance.py script works properly unless Anna tells me otherwise. And re: the gigahuge feature importance, your input sounds right, and they may need more training examples. And I was testing how the classifier performed by 'over-fitting', i.e. completely filling-in a few objects in a handful of images, and evaluating how the remaining objects in those same images were classified. I will expand my test set further. Thanks, David > Hi David, > > > (1) Should I trust this fairly old script? > Well, as far as I know, the way the variable importance is computed > inside vigra has not changed, even though the script is rather old. > @Anna: thanks for double checking!! > > however at the "gigahuge" scale all feature categories show quite high > > importance, even though I don't expect some of them, > For me this sound like a fundamental issue. Gigahuge features take a > (giga) huge neighborhood into account and one -in principle - needs more > examples to learn such structures (just because of combinatorials)... > Perhaps this is a side effect off over-fitting!? How does the classifier > generalize on other images? > > Cheers, > Christoph > -------------- next part -------------- An HTML attachment was scrubbed... URL: From anne.jorstad at epfl.ch Mon Apr 14 20:55:36 2014 From: anne.jorstad at epfl.ch (Jorstad Anne Alison) Date: Mon, 14 Apr 2014 18:55:36 +0000 Subject: [ilastik-user] Carving Preprocessing is Extremely Slow Message-ID: <10005FF4697F344388409544CE12986321BD6099@REXME.intranet.epfl.ch> Hi- I am trying to use the carving tool in Ilastik 1.0.5 to segment mitochondria in FIBSEM volumes, but the preprocessing computations are extremely slow. My cube is 1024x768x165, but I have zoomed in so am viewing about 25% of this in every dimension. I had understood that the previewing a smaller portion of the image stack would result in a smaller portion being processed, and the operations should therefore be faster. However, each new preprocessing calculation (changing the filter scale or changing the filter) takes many minutes to update, far slower than is practical for use. Even just zooming in or moving through the image takes tens of seconds to process. I am running Linux on a reasonably fast desktop machine, or alternately on a server with many processors. My question is, is there anything that can be done to parallelize these operations so that these processes can be sped up? Or is there anything else I can do to be able to interact with the stack in something closer to real time? Thank you for your help, I am excited about the future possibilities of this tool. --Anne From ullrich.koethe at iwr.uni-heidelberg.de Tue Apr 15 13:02:07 2014 From: ullrich.koethe at iwr.uni-heidelberg.de (Ullrich Koethe) Date: Tue, 15 Apr 2014 13:02:07 +0200 Subject: [ilastik-user] Carving Preprocessing is Extremely Slow In-Reply-To: <10005FF4697F344388409544CE12986321BD6099@REXME.intranet.epfl.ch> References: <10005FF4697F344388409544CE12986321BD6099@REXME.intranet.epfl.ch> Message-ID: <534D11AF.2030006@iwr.uni-heidelberg.de> Hi Anne, > I am trying to use the carving tool in Ilastik 1.0.5 to segment > mitochondria in FIBSEM volumes, but the preprocessing computations > are extremely slow. My cube is 1024x768x165, but I have zoomed in so > am viewing about 25% of this in every dimension. I had understood > that the previewing a smaller portion of the image stack would result > in a smaller portion being processed, and the operations should > therefore be faster. Ideally, this would be correct. In practice, carving preprocessing is not yet capable of processing just part of the data. Improving this is high on our priority list, but we have to implement a number of prerequisite improvements before we can actually go ahead. > My question is, is there anything that can be done to parallelize > these operations so that these processes can be sped up? Or is there > anything else I can do to be able to interact with the stack in > something closer to real time? For the time being, you can cut out a representative part of the data (it should be as small as possible) and write it into a separate file. Then the search for optimal preprocessing parameters should be done on just the small part. This way, you only have to preprocess the entire dataset once, using the final parameterization (note that the results of preprocessing are stored in the project file and don't have to be recomputed every time you open ilastik). If this is still too slow, you should copy ilastik's console output (which contains precise timings of the individual preprocessing steps) and send it to this list. This more detailed information should help identifying the bottleneck and give more specific advise. Best regards Ulli From corrado.cali at kaust.edu.sa Sun Apr 27 07:16:26 2014 From: corrado.cali at kaust.edu.sa (Corrado Cali) Date: Sun, 27 Apr 2014 05:16:26 +0000 Subject: [ilastik-user] Carving in version 1.0.5 for linux Message-ID: Hi all, we recently installed iLastik 1.0.5 in our linux systems. We are very happy about the way it handles bigger datasets, as well as the way the carving have been modified. Actually, I could test the carving module only on my personal mac machine (1.0.5c on OSx mavericks), but trouble is that there is no button to export the mesh. In the linux version, the carving module seems missing at all, so we managed only to load the images. Anyone reported the same problem? Any chance of fixing it anytime soon? Thanks, CC :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: Ing. Corrado Cal?, PhD Post Doctoral Fellow Laboratory of Cell Imaging and Energetics Division of BESE Room 4266, Level 4, Building 2 4700 King Abdullah University of Science and Technology Thuwal 23955-6900, Kingdom of Saudi Arabia Tel: +966 (0)2 808 2969 skype: menestrello82 Please note that week ends on the kingdom are on fridays and saturdays :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. -------------- next part -------------- An HTML attachment was scrubbed... URL: From ji.tang at zeiss.com Fri Aug 1 09:31:28 2014 From: ji.tang at zeiss.com (Tang, Ji) Date: Fri, 1 Aug 2014 07:31:28 +0000 Subject: [ilastik-user] Add label Message-ID: <796CB20F6C5F534B900EAE7BB1C3C17A14581B@ADEERL01SMS001.cznet.zeiss.org> Dear, I encountered a problem when I used ilastik, so I look for help. The problem is that I often cannot add label to objects when I do object classification. I try to solve it by searching, but little information can be found about ilastik usage. By the way, I don't understand "Two thresholds". Could you help me? Thank you very much. -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Fri Aug 1 12:05:28 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Fri, 01 Aug 2014 12:05:28 +0200 Subject: [ilastik-user] Add label In-Reply-To: <796CB20F6C5F534B900EAE7BB1C3C17A14581B@ADEERL01SMS001.cznet.zeiss.org> References: <796CB20F6C5F534B900EAE7BB1C3C17A14581B@ADEERL01SMS001.cznet.zeiss.org> Message-ID: <53DB6668.9030001@iwr.uni-heidelberg.de> Hello, Could you describe in a bit more detail what you are trying to do? Do I get it right that you are trying to click on objects and they don't get labeled? Only for some objects or all of them? Adding labels in the left-side menu went ok? Can you try to right-click on the object and select "Label with Label 1" from the context menu? Does that work? Can you send us the .ilastik_log.txt file, which should be located in your home directory? Two-level thresholding is a bit more tricky. Essentially, it performs hysteresis thresholding with a size filter on probability maps, produced by the pixel classification workflow. In short, it first thresholds at a high threshold, say 0.95 and thus selects only the regions of very high probability. Then it performs the size filtering. As the last step, it takes the regions which passed the first threshold and the size filter and relaxes the threshold to the low threshold value. This way, you can get whole objects instead of only the cores, which pass the high threshold, and still get some filtering done. Does it make sense now? Cheers, Anna On 01.08.2014 09:31, Tang, Ji wrote: > > Dear, > > I encountered a problem when I used ilastik, so I look for help. The > problem is that I often cannot add label to objects when I do object > classification. I try to solve it by searching, but little information > can be found about ilastik usage. > > By the way, I don't understand "Two thresholds". > > Could you help me? Thank you very much. > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Fri Aug 1 13:39:38 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Fri, 01 Aug 2014 13:39:38 +0200 Subject: [ilastik-user] Problems with batch processing in cell density counting In-Reply-To: <85299086-F2B2-4314-8C1F-8D4FE15C94CC@nimr.mrc.ac.uk> References: <85299086-F2B2-4314-8C1F-8D4FE15C94CC@nimr.mrc.ac.uk> Message-ID: <53DB7C7A.5030402@iwr.uni-heidelberg.de> Dear Anna, Thanks a lot for pointing out this problem! I created an issue on the ilastik issue tracker (https://github.com/ilastik/ilastik/issues/991) and we'll do our best to fix it as soon as possible. I'll let you know as soon as it's done, Anna On 31.07.2014 17:00, Anna Olerinyova wrote: > Good afternoon, > > I am currently trying to use Ilastik 1.1.0 to enumerate granule cells > in cerebellum, but when I get to the batch processing, the Batch > Prediction Output Location doesn't seem to work at all. Regardless of > whether I put in one or may images into the Input Selection, there is > no outcome and I cannot export them at all. Any ideas what might be wrong? > > Many thanks, > Anna Olerinyova > > > ------------------------------------------------------------------------ > This email has been scanned for email related threats and delivered > safely by Mimecast. > For more information please visit http://www.mimecast.com > ------------------------------------------------------------------------ > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From bergs at janelia.hhmi.org Fri Aug 1 17:00:38 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Fri, 1 Aug 2014 15:00:38 +0000 Subject: [ilastik-user] Problems with batch processing in cell density counting In-Reply-To: <53DB7C7A.5030402@iwr.uni-heidelberg.de> References: <85299086-F2B2-4314-8C1F-8D4FE15C94CC@nimr.mrc.ac.uk> <53DB7C7A.5030402@iwr.uni-heidelberg.de> Message-ID: An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Tue Aug 5 13:31:12 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Tue, 05 Aug 2014 13:31:12 +0200 Subject: [ilastik-user] =?gb2312?b?tPC4tDogIEFkZCBsYWJlbA==?= In-Reply-To: <796CB20F6C5F534B900EAE7BB1C3C17A145911@ADEERL01SMS001.cznet.zeiss.org> References: <796CB20F6C5F534B900EAE7BB1C3C17A14581B@ADEERL01SMS001.cznet.zeiss.org> <53DB6668.9030001@iwr.uni-heidelberg.de> <796CB20F6C5F534B900EAE7BB1C3C17A145911@ADEERL01SMS001.cznet.zeiss.org> Message-ID: <53E0C080.5080507@iwr.uni-heidelberg.de> Dear Tang Ji, By your description it sounds like you thresholded the background class. Could you switch on the "Objects" layer in the lower left part of the ilastik screen and check if your objects appear in different colors? If not, try going back to thresholding and check, at the top of the thresholding applet, if the you selected the foreground channel. Also, look at the "Final Output" layer, do the objects appear in different colors there? Hope it helps, Anna On 05.08.2014 10:56, Tang, Ji wrote: > > Dear Anna, > > Thank you very much for your detailed reply. However, I cannot still > add label. I am trying to do object classification. But after I click > ?Add Label? and then click some of objects, there is no respond. All > objects cannot be labeled except for the background. Adding labels in > the left-side menu doesn?t work. It doesn?t work too when I try to > right-click on the object and select "Label with Label 1" from the > context menu. I don?t find the ?.ilastik_log.txt? file after I search > for it. > > Thank you very much. > > Cheers, > > Tang Ji > > *???**:*Anna Kreshuk [mailto:anna.kreshuk at iwr.uni-heidelberg.de] > *????:* 2014?8?1? 18:05 > *???:* Tang, Ji; ilastik-user at ilastik.org > *??:* Re: [ilastik-user] Add label > > Hello, > > Could you describe in a bit more detail what you are trying to do? Do > I get it right that you are trying to click on objects and they don't > get labeled? Only for some objects or all of them? Adding labels in > the left-side menu went ok? Can you try to right-click on the object > and select "Label with Label 1" from the context menu? Does that work? > Can you send us the .ilastik_log.txt file, which should be located in > your home directory? > > Two-level thresholding is a bit more tricky. Essentially, it performs > hysteresis thresholding with a size filter on probability maps, > produced by the pixel classification workflow. In short, it first > thresholds at a high threshold, say 0.95 and thus selects only the > regions of very high probability. Then it performs the size filtering. > As the last step, it takes the regions which passed the first > threshold and the size filter and relaxes the threshold to the low > threshold value. This way, you can get whole objects instead of only > the cores, which pass the high threshold, and still get some filtering > done. Does it make sense now? > > Cheers, > Anna > > On 01.08.2014 09:31, Tang, Ji wrote: > > Dear, > > I encountered a problem when I used ilastik, so I look for help. > The problem is that I often cannot add label to objects when I do > object classification. I try to solve it by searching, but little > information can be found about ilastik usage. > > By the way, I don?t understand ?Two thresholds?. > > Could you help me? Thank you very much. > > > > > _______________________________________________ > > ilastik-user mailing list > > ilastik-user at ilastik.org > > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > -------------- next part -------------- An HTML attachment was scrubbed... URL: From d.pflugfelder at fz-juelich.de Wed Aug 27 11:55:57 2014 From: d.pflugfelder at fz-juelich.de (Pflugfelder, Daniel) Date: Wed, 27 Aug 2014 09:55:57 +0000 Subject: [ilastik-user] 3dview Message-ID: <2ECFDD50D27D624C8055143F6D7FF97F2910830B@MBX2010-E01.ad.fz-juelich.de> Dear all, I just started using ilastik (version 1.1) and I do not get the 3dview to work. I have a 3d data set and used Ilastik to generate nice segmentations with the pixel classification module, however I only see the segmentation on the 2d cutting plant. How can I display the data in the attached 3d-viewer? Cheers, Daniel ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------ Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------ -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Wed Aug 27 21:19:37 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Wed, 27 Aug 2014 21:19:37 +0200 Subject: [ilastik-user] 3dview In-Reply-To: <2ECFDD50D27D624C8055143F6D7FF97F2910830B@MBX2010-E01.ad.fz-juelich.de> References: <2ECFDD50D27D624C8055143F6D7FF97F2910830B@MBX2010-E01.ad.fz-juelich.de> Message-ID: <20140827211937.o0j6i1ywbo08s8wo@wwwmail.urz.uni-heidelberg.de> Dear Daniel, Thanks a lot for your feedback, it's very important for us to know which features users miss the most. The 3d view is only activated in the carving workflow, not pixel classification. The reason for this is that pixel classification no longer assumes that the data fits in memory and if the data is too large we can't run the mesh generation on the whole block. Doing it just for the field of view is on our roadmap, though, so this feature will become available, but it's hard to say when exactly. For now, I'm afraid, you'll have to export your segmentation results to another tool to do 3D visualization. Cheers, Anna Quoting "Pflugfelder, Daniel" : > Dear all, > > I just started using ilastik (version 1.1) and I do not get the > 3dview to work. > I have a 3d data set and used Ilastik to generate nice segmentations > with the pixel classification module, however I only see the > segmentation on the 2d cutting plant. > How can I display the data in the attached 3d-viewer? > > Cheers, > Daniel > > > > ------------------------------------------------------------------------------------------------ > ------------------------------------------------------------------------------------------------ > Forschungszentrum Juelich GmbH > 52425 Juelich > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 > Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher > Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), > Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, > Prof. Dr. Sebastian M. Schmidt > ------------------------------------------------------------------------------------------------ > ------------------------------------------------------------------------------------------------ > > From messages-noreply at linkedin.com Mon Dec 29 10:03:10 2014 From: messages-noreply at linkedin.com (Dr. Kumar Rajamani) Date: Mon, 29 Dec 2014 09:03:10 +0000 (UTC) Subject: [ilastik-user] I'd like to add you to my professional network on LinkedIn Message-ID: <820189690.11587562.1419843790259.JavaMail.app@lva1-app2741.prod> Hi, I'd like to add you to my professional network on LinkedIn. - Dr. Kumar Accept: https://www.linkedin.com/blink?simpleRedirect=0ScjgOdjgTc3kUcjgSdj8RdjARfkh9rCZFt65QqnpKqip2szlMtzhAgzRVpkJApn9xq7cCt7dBtmtvpnhFtCVFfmJB9C5QoORBt6BSrCAJt7dBtmsJr6RBfmtKqmJzon9Q9ClQqnpKimtBkClOs3Rx9zoSbndSsm9JejhFbm4RdmpCrjRBfP9SbSkLrmZzbCVFp6lHrCBIbDtTtOYLeDdMt7hE&msgID=I5955256437633478656_500&markAsRead= You received an invitation to connect. LinkedIn will use your email address to make suggestions to our members in features like People You May Know. Unsubscribe here: https://www.linkedin.com/blink?simpleRedirect=0SdyRPtD5yrjAQqiRxdjlCpCQZp6BB9DsVdTFhpPt5hCtdlkFfnQt1tD9zgk9HmDtprD1DclpHpREMcQlqmAtcgzxPe6d2d5Fypnl6dBBnpSFOdSNHdRpfkRpNlk51gk4RpTlRdkxos7sThB51fmhFrSMCt7dBtmtvpnhFtCVFfmJB9B4ScSh5hSxmhD9JhB51fmVBqSZkp6BJ9CVRr3RQ9zoSbndSsm9JejhFbm4RdmpCrjRBfP9SbSkLrmZzbCVFp6lHrCBIbDtTtOYLeDdMt7hE&msgID=I5955256437633478656_500&markAsRead= Learn why we included this at the following link: https://www.linkedin.com/blink?simpleRedirect=e3wTd3RAimlIoSBQsC4Ct7dBtmtvpnhFtCVFfmJB9CNOlmlzqnpOpldOpmRLt7dRoPRx9zoSbndSsm9JejhFbm4RdmpCrjRBfP9SbSkLrmZzbCVFp6lHrCBIbDtTtOYLeDdMt7hE&msgID=I5955256437633478656_500&markAsRead= © 2014, LinkedIn Corporation. 2029 Stierlin Ct. Mountain View, CA 94043, USA -------------- next part -------------- An HTML attachment was scrubbed... URL: From Ana.M.C.Beltran at ww.uni-erlangen.de Wed Feb 26 17:58:29 2014 From: Ana.M.C.Beltran at ww.uni-erlangen.de (Ana.M.C.Beltran at ww.uni-erlangen.de) Date: Wed, 26 Feb 2014 17:58:29 +0100 Subject: [ilastik-user] memory requirements Message-ID: <0c303e2f36d827f251324ec2940c278d.squirrel@faumail.uni-erlangen.de> Dear All, Is it a special memory recquired to run Ilastik 0.5.012? or is there any limit for the files to use? is there any settings I can modify to work quicker? I am trying to segment tomography data and for files around 300 MB the software doesn?t work, and the computer have a big memory. Could you give me some ideas? Thank you very much in advance best regards - Dr.Ing. Ana M Beltran From christoph.sommer at imba.oeaw.ac.at Thu Feb 27 16:14:35 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Thu, 27 Feb 2014 16:14:35 +0100 Subject: [ilastik-user] memory requirements In-Reply-To: <0c303e2f36d827f251324ec2940c278d.squirrel@faumail.uni-erlangen.de> References: <0c303e2f36d827f251324ec2940c278d.squirrel@faumail.uni-erlangen.de> Message-ID: <530F565B.1090605@imba.oeaw.ac.at> Dear Ana, the binarz ilastik 0.5.012 is compiled for 64bit, so the sheer memory should not be the issue. Can you load the data? What format is your data it stroed in? Is there any error message? Best, Christoph > Dear All, > > Is it a special memory recquired to run Ilastik 0.5.012? or is there any > limit for the files to use? is there any settings I can modify to work > quicker? > I am trying to segment tomography data and for files around 300 MB the > software doesn?t work, and the computer have a big memory. > Could you give me some ideas? > Thank you very much in advance > best regards > > - > Dr.Ing. Ana M Beltran > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From sommer.chr at gmx.de Thu Feb 27 16:14:54 2014 From: sommer.chr at gmx.de (Christoph Sommer) Date: Thu, 27 Feb 2014 16:14:54 +0100 Subject: [ilastik-user] memory requirements In-Reply-To: <0c303e2f36d827f251324ec2940c278d.squirrel@faumail.uni-erlangen.de> References: <0c303e2f36d827f251324ec2940c278d.squirrel@faumail.uni-erlangen.de> Message-ID: <530F566E.2060100@gmx.de> Dear Ana, the binarz ilastik 0.5.012 is compiled for 64bit, so the sheer memory should not be the issue. Can you load the data? What format is your data it stroed in? Is there any error message? Best, Christoph > Dear All, > > Is it a special memory recquired to run Ilastik 0.5.012? or is there any > limit for the files to use? is there any settings I can modify to work > quicker? > I am trying to segment tomography data and for files around 300 MB the > software doesn?t work, and the computer have a big memory. > Could you give me some ideas? > Thank you very much in advance > best regards > > - > Dr.Ing. Ana M Beltran > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From evelzw at gmail.com Wed Jan 1 16:48:40 2014 From: evelzw at gmail.com (Zhaowen Liu) Date: Wed, 1 Jan 2014 10:48:40 -0500 Subject: [ilastik-user] Advice on identifying different particle shapes Message-ID: <24C57499-82E3-4F8F-80D1-6F95BA730ECC@gmail.com> Hi everyone, I have batches of SEM images of gold nano particles. Particles are formed in different shapes, singlets, dimers, trimers and big clusters etc. I would like to identify them with Ilsatik. The dark part of the image is background and bright objects are nano particles. I tried Classification in Ilastic with 5 color labels, red for background, green for singlets, yellow for dimers, blue for trimers and magenta for big clusters. But according to the Segmentation result, they can not be separated form each other no matter how many labels i added. So i am wondering if it is possible to classify different object types of similar color with Ilastik. Or if it is not, do you have any suggestion for some other methods to identify singlets, dimers and clusters etc? To illustrate what different shapes look like, I attached the original SEM image, the diagram of shapes and the Segmentation output from Ilastik as follows. Thanks a lot for your time and happy new year! Best, Evelyn -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: particle_types.tiff Type: image/tiff Size: 732862 bytes Desc: not available URL: From neuhausmj at gmail.com Fri Jan 3 17:58:41 2014 From: neuhausmj at gmail.com (Miranda Segura) Date: Fri, 3 Jan 2014 10:58:41 -0600 Subject: [ilastik-user] Uploading 3D files other than images and stacks In-Reply-To: <46f75899d6404eb79d538d623bf6bd21@CHIHT2.ad.uillinois.edu> References: <46f75899d6404eb79d538d623bf6bd21@CHIHT2.ad.uillinois.edu> Message-ID: Hello, I'm writing again to ask about uploading 3D files that are not .h5 format. I am able to save the files from my confocal scope as .txt, .stl, .set, .sur, .map, and a few more, but I can not save the files in .h5 format. Is it absolutely necessary to convert these files to .h5? Thanks for any help! Miranda On Tue, Nov 26, 2013 at 9:26 AM, Miranda Segura wrote: > Hello, > I am using a Nanofocus confocal 3D measurement system to generate > thousands of leaf surface topographies. In addition to the topography > file, the software also outputs a reflective intensity file (what our eyes > see). From these measurements, I want to categorize two cell types, so > Ilastik seems like it could do the trick- if I could upload the files. > > The files that are generated from this software include an .stl or .set > file format. They can also generate .mnt or .nms files, but those are > specific to Nanofocus software (I believe). > > Is there anyway I can upload and use other file types in Ilastik? > Thanks for any help! > > -- > Miranda Segura > PhD Candidate > Dept. of Plant Biology > University of Illinois-Champaign Urbana > > -- Miranda Segura PhD Candidate Dept. of Plant Biology University of Illinois -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Tue Jan 7 14:58:52 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Tue, 07 Jan 2014 14:58:52 +0100 Subject: [ilastik-user] Uploading 3D files other than images and stacks In-Reply-To: References: <46f75899d6404eb79d538d623bf6bd21@CHIHT2.ad.uillinois.edu> Message-ID: <52CC081C.7070404@iwr.uni-heidelberg.de> Hi Miranda, I don't think we can read any of the formats you mention, but if you can save the images as a stack of tifs or other image formats, we can import it as 3D (choose the "Load stack" option for that, when loading data). For example, if FiJi can load your data, there is a "save as image sequence" option there, which writes out files as we can read them. Hope it helps, Anna On 03.01.2014 17:58, Miranda Segura wrote: > Hello, > I'm writing again to ask about uploading 3D files that are not .h5 > format. I am able to save the files from my confocal scope as .txt, > .stl, .set, .sur, .map, and a few more, but I can not save the files > in .h5 format. Is it absolutely necessary to convert these files to .h5? > > Thanks for any help! > > Miranda > > > On Tue, Nov 26, 2013 at 9:26 AM, Miranda Segura > wrote: > > Hello, > I am using a Nanofocus confocal 3D measurement system to > generate thousands of leaf surface topographies. In addition to > the topography file, the software also outputs a reflective > intensity file (what our eyes see). From these measurements, I > want to categorize two cell types, so Ilastik seems like it could > do the trick- if I could upload the files. > > The files that are generated from this software include an .stl > or .set file format. They can also generate .mnt or .nms files, > but those are specific to Nanofocus software (I believe). > > Is there anyway I can upload and use other file types in > Ilastik? Thanks for any help! > > -- > Miranda Segura > PhD Candidate > Dept. of Plant Biology > University of Illinois-Champaign Urbana > > > > > -- > Miranda Segura > PhD Candidate > Dept. of Plant Biology > University of Illinois > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From corrado.cali at kaust.edu.sa Mon Jan 20 18:17:08 2014 From: corrado.cali at kaust.edu.sa (Corrado Cali) Date: Mon, 20 Jan 2014 17:17:08 +0000 Subject: [ilastik-user] Trouble with spatial features Message-ID: <66BA840C-E855-4A54-A047-92EAD50CBF3F@kaust.edu.sa> Hi all, I am experience a very strange error in iLastik lately. I am running iLastik 0.5.12 on a MacBookPro , 16GB DDR3. Lately, after I load an EM stack, I go on the Classification window, then I click on Select Features, to select the spatial features. At these point, usually I could pass with the mouse on the different sizes, and a red circle with the corresponding size appeared on a preview of the image on the box on the right. What is happening now is that when I pass with the mouse on whatever feature, nothing appears, and the software enters in a loop, it keeps loading and I can?t go on doing anything. I never experienced that, and the only thing I can think about is that I updated my OSx to mavericks, the latest release, few days ago. Have anyone of you experienced such an issue? This is very odd, I never had problems at this point. And this machine is quite powerful, I already segmented the same stack using iLastik before. Thanks everyone in advance, best regards. CC :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: Ing. Corrado Cal?, PhD Post Doctoral Fellow Laboratory of Cell Imaging and Energetics Division of BESE Room 4266, Level 4, Building 2 4700 King Abdullah University of Science and Technology Thuwal 23955-6900, Kingdom of Saudi Arabia Tel: +966 (0)2 808 2969 skype: menestrello82 Please note that week ends on the kingdom are on fridays and saturdays :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. -------------- next part -------------- An HTML attachment was scrubbed... URL: From mailfert at ciml.univ-mrs.fr Thu Jul 10 09:52:58 2014 From: mailfert at ciml.univ-mrs.fr (Sebastien MAILFERT) Date: Thu, 10 Jul 2014 07:52:58 +0000 Subject: [ilastik-user] Problem with Windows 8 32 bits installation Message-ID: Dear users, I'm a newcomer to this forum. I have a colleagues who asks me to install ilastik on a Windows8 32 bits computer. The installation seems to be OK but when we launch the executable, it seems that the program is not compatible with this OS. We tried both installer: Windows XP/Vista/7 32 bit installer [*] ilastik-0.5-setup-v0.5.06.rc5-x86-32.exe [*] ilastik-0.4-setup-rev553.exe Do you have a idea to help us please? Best regards. S?bastien S?bastien MAILFERT Research engineer, PhD ImagImm : Imaging Immunity Photonic microscopy facility Tel: 33 (0)4 91 26 91 26 [signature_mail_CIML_2011] -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.jpg Type: image/jpeg Size: 3898 bytes Desc: image001.jpg URL: From joshua.skootsky at gmail.com Mon Jul 14 17:15:39 2014 From: joshua.skootsky at gmail.com (Joshua Skootsky) Date: Mon, 14 Jul 2014 11:15:39 -0400 Subject: [ilastik-user] 5D data input into ilastik Message-ID: Hello, I can see that there are issues with loading 5D data into Ilastik https://github.com/ilastik/ilastik/issues/867 http://burgerdev.de/ I'm trying to load the GoFigure sample data into Ilastik ( http://www.gofigure2.org/index.php/users.html - see GoFigureSampleData.zip) The problem is that Ilastik doesn't recognize it as 3d+t data, and doesn't let me change the channels afterwards. How can this be done? -- Joshua Skootsky (415) 683-9354 | joshua.skootsky at gmail.com -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Fri Jul 25 23:25:18 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Fri, 25 Jul 2014 23:25:18 +0200 Subject: [ilastik-user] 5D data input into ilastik In-Reply-To: References: Message-ID: <20140725232518.fvvvi0un4gkc0gsk@wwwmail.urz.uni-heidelberg.de> Hi Joshua, I just tried loading a stack, made of several pngs from the GoFigure example, and changing the axis order to tyxc. It worked fine for me in the Pixel Classification workflow. I guess I'm not doing exactly what you are trying to do. How are you loading the data? How are you trying to change the channels? Which project are you using? Thanks for your feedback Anna Quoting Joshua Skootsky : > Hello, > > I can see that there are issues with loading 5D data into Ilastik > > https://github.com/ilastik/ilastik/issues/867 > > http://burgerdev.de/ > > I'm trying to load the GoFigure sample data into Ilastik ( > http://www.gofigure2.org/index.php/users.html - see GoFigureSampleData.zip) > > The problem is that Ilastik doesn't recognize it as 3d+t data, and doesn't > let me change the channels afterwards. > > How can this be done? > > > -- > Joshua Skootsky > (415) 683-9354 | joshua.skootsky at gmail.com > From jgburchfield.au at gmail.com Mon Jul 28 08:29:06 2014 From: jgburchfield.au at gmail.com (James Burchfield) Date: Mon, 28 Jul 2014 16:29:06 +1000 Subject: [ilastik-user] Questions related to RGB image segmentation Message-ID: Hi, I am using ilastik on a set of RGB images. The orginal images are composite-tiff files. When loaded into ilastik they open, but the channels are interpreted as z. I cannot work out how to change this in ilastik. I tried editing the metadata, but this files. Conversion to RGB outside of ilastik solves the problem, however, I would like to be able to view the channels independently within ilastik. Ideally change individual channel brightness. I think this may have been possible in 0.5......... This would make training significantly easier. Also, I ultimately plan to use more than 3 channels of data. Is this possible? Cheers, James -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Tue Jul 29 15:29:19 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Tue, 29 Jul 2014 15:29:19 +0200 Subject: [ilastik-user] Questions related to RGB image segmentation In-Reply-To: References: Message-ID: <53D7A1AF.5000806@iwr.uni-heidelberg.de> Dear James, We are aware of this problem and we are working on fixing it. In general, if you have an RGB image loaded as RGB, that is, when ilastik does not mistake it for a z-stack, there is an option in the Input dialog to display the channels individually (screenshot attached, select "Grayscale" from the pointed drop-down menu). Using more than 3 channels is also no problem, as long as ilastik does not interpret them as a z-stack. Sorry for the inconvenience, we'll let you know as soon as it's fixed Anna On 28.07.2014 08:29, James Burchfield wrote: > Hi, > I am using ilastik on a set of RGB images. The orginal images are > composite-tiff files. > When loaded into ilastik they open, but the channels are interpreted as z. > I cannot work out how to change this in ilastik. I tried editing the > metadata, but this files. > Conversion to RGB outside of ilastik solves the problem, however, I > would like to be able to view the channels independently within > ilastik. Ideally change individual channel brightness. I think this > may have been possible in 0.5......... > This would make training significantly easier. Also, I ultimately > plan to use more than 3 channels of data. Is this possible? > > Cheers, > > James > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: DatasetPropertiesDialog.png Type: image/png Size: 44683 bytes Desc: not available URL: From anna.kreshuk at iwr.uni-heidelberg.de Wed Jul 30 11:40:28 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Wed, 30 Jul 2014 11:40:28 +0200 Subject: [ilastik-user] Questions related to RGB image segmentation In-Reply-To: References: <53D7A1AF.5000806@iwr.uni-heidelberg.de> Message-ID: <53D8BD8C.2000301@iwr.uni-heidelberg.de> Hi James, Thanks for pointing it out, we had an error message pop out too early in this case. I changed it so that you can now save the the multi-channel outputs as an image sequence. We are planning to do a bug-fix release soon (hopefully, this week) and this fix will be in there. Until the release is out, you have to either save .h5 files and convert them to images later, or, with a bit more clicking, you can save individual channels as images. To do that, get the "Export settings dialog" and limit the channel range to 1 (or up to 3) channels of interest. You can then save the selected channels in any image format (see attached screenshot). The settings apply to all images in your batch input list, so you don't have to repeat it every time. Hope this helps, Anna On 30.07.2014 07:58, James Burchfield wrote: > Hi Anna, > Another question. > Exporting the data. What is the best way to export the .h5 files? > None of the alternative file types seem to like having 5 > channels............ > Cheers, > James > -------------- next part -------------- A non-text attachment was scrubbed... Name: Image_Export_Options_748.png Type: image/png Size: 80230 bytes Desc: not available URL: From aolerin at nimr.mrc.ac.uk Thu Jul 31 17:00:11 2014 From: aolerin at nimr.mrc.ac.uk (Anna Olerinyova) Date: Thu, 31 Jul 2014 16:00:11 +0100 Subject: [ilastik-user] Problems with batch processing in cell density counting Message-ID: <85299086-F2B2-4314-8C1F-8D4FE15C94CC@nimr.mrc.ac.uk> Good afternoon, I am currently trying to use Ilastik 1.1.0 to enumerate granule cells in cerebellum, but when I get to the batch processing, the Batch Prediction Output Location doesn't seem to work at all. Regardless of whether I put in one or may images into the Input Selection, there is no outcome and I cannot export them at all. Any ideas what might be wrong? Many thanks, Anna Olerinyova --------------------------------------------------------------------------------------- This email has been scanned for email related threats and delivered safely by Mimecast. For more information please visit http://www.mimecast.com --------------------------------------------------------------------------------------- -------------- next part -------------- An HTML attachment was scrubbed... URL: From raphael at aims.ac.za Mon Jun 2 22:35:33 2014 From: raphael at aims.ac.za (Raphael Okoye) Date: Mon, 2 Jun 2014 13:35:33 -0700 Subject: [ilastik-user] cell profiler pipeline Message-ID: hi all, I have 300 images I want to analyze using the ilastik cell profiler module. I selected some of the images and used them in training and exported the .h5 classifier file. How can I use this in the cell profiler pipeline to analyze my 300 images? The example on http://www.ilastik.org/index.php?cat=20_Documentation&page=03_Cellprofiler is just for 1 image. Thanks as I look forward to hearing from you guys. Regards, Raphael. -------------- next part -------------- An HTML attachment was scrubbed... URL: From christoph.sommer at imba.oeaw.ac.at Wed Jun 4 11:28:43 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Wed, 4 Jun 2014 11:28:43 +0200 Subject: [ilastik-user] cell profiler pipeline In-Reply-To: References: Message-ID: <538EE6CB.4080600@imba.oeaw.ac.at> Hi Raphael, this depends on on your specific CellProfiler Pipeline. the ilastik module "ClassifyPixels" behaves just as any other CP module. Have a look here for examples how to set up CellProfiler pipelines that involve processing of more images. http://www.cellprofiler.org/examples.shtml Cheers, Christoph > hi all, > > I have 300 images I want to analyze using the ilastik cell profiler > module. I selected some of the images and used them in training and > exported the .h5 classifier file. How can I use this in the cell > profiler pipeline to analyze my 300 images? The example on > http://www.ilastik.org/index.php?cat=20_Documentation&page=03_Cellprofiler > is just for 1 image. Thanks as I look forward to hearing from you guys. > > Regards, > Raphael. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From andrew.nere at gmail.com Tue Jun 17 09:15:12 2014 From: andrew.nere at gmail.com (Andrew Nere) Date: Tue, 17 Jun 2014 07:15:12 +0000 Subject: [ilastik-user] (no subject) Message-ID: <29FF0B35-771C-4E99-B849-58458CE24212@atarde.com.br> An embedded and charset-unspecified text was scrubbed... Name: not available URL: -------------- next part -------------- An embedded message was scrubbed... From: "Andrew Nere" Subject: Date: Tue, 17 Jun 2014 07:15:12 +0000 Size: 1879 URL: From glendon.holst at kaust.edu.sa Sun Jun 22 12:48:09 2014 From: glendon.holst at kaust.edu.sa (Glendon Holst) Date: Sun, 22 Jun 2014 13:48:09 +0300 Subject: [ilastik-user] ilastik asserts error when changing Preprocessing values Message-ID: <53A6B469.8050802@kaust.edu.sa> I encountered errors after changing the values on the Preprocessing tab of the Carving workflow. I had labeled and segmented two object already, saved them and the entire project, then changed "dark lines (inverted, smoothed)" to "dark lines (valley)" and the "filter scale" value. I had to uncheck "write protect". I had expected that the original labels would be preserved, but the exact segmentation lost (just need to recreate it). I got the following errors ("IndexError: index 266034 is out of bounds for size 259728") when selecting the Labelling tab. I was able to re-Segment the two objects and get rid of those errors. (After saving the objects I was also able to save the project without error). Would it be possible to remove the computed values (segmentations) dependent on the old Preprocessing values (but keep the labels) to prevent the errors? Thanks! Glendon ERROR 2014-06-22 10:42:31,414 excepthooks 18956 140053465376512 Traceback (most recent call last): File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/imageScene2D.py", line 390, in drawBackground for tile in tiles: File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 534, in getTiles self.requestRefresh( rectF ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 556, in requestRefresh self._refreshTile( stack_id, tile_no ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 609, in _refreshTile img = ims_req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 377, in wait return self.toImage() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 383, in toImage self._arrayreq.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/slicesources.py", line 43, in wait return self._sp(self._ar.wait()) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 444, in wait rawData = self._rawRequest.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 207, in wait a = self._req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/operators/opReorderAxes.py", line 114, in execute self.Input( *in_roi ).writeInto( result_input_view ).wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 619, in __call__ return self.func( *totalargs, **self.kwargs) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/network/fas6280-a.vis.kaust.edu.sa/vol/projects/epfl/software/ilastik/1.0.6/src/ilastik/ilastik/ilastik/workflows/carving/opCarving.py", line 636, in execute temp = self._done_seg_lut[self._mst.regionVol[sl[1:4]]] IndexError: index 266034 is out of bounds for size 259728 ERROR 2014-06-22 10:42:31,447 excepthooks 18956 140053465376512 Traceback (most recent call last): File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/imageScene2D.py", line 390, in drawBackground for tile in tiles: File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 534, in getTiles self.requestRefresh( rectF ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 556, in requestRefresh self._refreshTile( stack_id, tile_no ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 609, in _refreshTile img = ims_req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 377, in wait return self.toImage() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 383, in toImage self._arrayreq.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/slicesources.py", line 43, in wait return self._sp(self._ar.wait()) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 444, in wait rawData = self._rawRequest.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 207, in wait a = self._req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/operators/opReorderAxes.py", line 114, in execute self.Input( *in_roi ).writeInto( result_input_view ).wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 619, in __call__ return self.func( *totalargs, **self.kwargs) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/network/fas6280-a.vis.kaust.edu.sa/vol/projects/epfl/software/ilastik/1.0.6/src/ilastik/ilastik/ilastik/workflows/carving/opCarving.py", line 636, in execute temp = self._done_seg_lut[self._mst.regionVol[sl[1:4]]] IndexError: index 260951 is out of bounds for size 259728 ERROR 2014-06-22 10:42:31,488 excepthooks 18956 140053465376512 Traceback (most recent call last): File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/imageScene2D.py", line 390, in drawBackground for tile in tiles: File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 534, in getTiles self.requestRefresh( rectF ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 556, in requestRefresh self._refreshTile( stack_id, tile_no ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/tiling.py", line 609, in _refreshTile img = ims_req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 377, in wait return self.toImage() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/imagesources.py", line 383, in toImage self._arrayreq.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/slicesources.py", line 43, in wait return self._sp(self._ar.wait()) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 444, in wait rawData = self._rawRequest.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/volumina/volumina/pixelpipeline/datasources.py", line 207, in wait a = self._req.wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/operators/opReorderAxes.py", line 114, in execute self.Input( *in_roi ).writeInto( result_input_view ).wait() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 333, in wait return self._wait(timeout) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 356, in _wait self._wait_within_foreign_thread( timeout ) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 240, in _execute self._result = self.fn() File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/request/request.py", line 619, in __call__ return self.func( *totalargs, **self.kwargs) File "/var/remote/projects/epfl/software/ilastik/1.0.6/src/ilastik/lazyflow/lazyflow/slot.py", line 770, in __call__ result_op = self.operator.execute(self.slot, (), self.roi, destination) File "/network/fas6280-a.vis.kaust.edu.sa/vol/projects/epfl/software/ilastik/1.0.6/src/ilastik/ilastik/ilastik/workflows/carving/opCarving.py", line 636, in execute temp = self._done_seg_lut[self._mst.regionVol[sl[1:4]]] IndexError: index 267089 is out of bounds for size 259728 ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. From christoph.sommer at imba.oeaw.ac.at Mon Mar 3 11:54:04 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 3 Mar 2014 11:54:04 +0100 Subject: [ilastik-user] [Fwd: memory requirements] In-Reply-To: <5af6b9214448ff1717a15885f8b0f14a.squirrel@faumail.uni-erlangen.de> References: <5af6b9214448ff1717a15885f8b0f14a.squirrel@faumail.uni-erlangen.de> Message-ID: <53145F4C.50008@imba.oeaw.ac.at> Dear Ana, what is the dimensionality of your data in pixels (x,y,z,c)? Is there also a temporal channel? ilastik 0.5 is still computing all outputs on the full image. For large images this can take very long, even if the image data fits into main memory. What people typically do, is to train the classifier on smaller subsets of the image or to downsample the data (if this is possible without loosing important information for the classification task at hand). Once a classifier is trained you can apply the it to bigger data in batch mode, which can run in the background (via Automate). Also note, that not only the image data itself has to fit into main memory, but also the image features and several intermediate results have to be allocated. Hence, using a large feature set, might lead to memory errors. Best, Christoph > Dear Christoph > > Today, I can report it says: Memory error, working with an original file > of 184 MB It uses the 27GB of the computer. > Please, could you confirm me the memory requirements of the software? I am > very interesting on using it for my research. > > Thank you very much on advance > > Best regards > > Dr. Ing. Ana M Beltr?n > > > Ana.M.C.Beltran at ww.uni-erlangen.de wrote: >> Dear Christoph >> >> I can load the data. I load series of images beause I want to do 3D >> segmentation. When the total file is around 60 MB, it works slow but >> without problems. However, I cannot work with bigger files. >> I am using a computer with 24GB RAM, and the memory is completely >> saturated. It works in Windows 7, should I change to windows XP or another >> version? >> The file is a series of tiff images, 8 bits, gray scale >> >> Thank you very much for your support >> >> Best regards >> >> Ana M Beltr?n >> >> >> ---------------------------- Original Message ---------------------------- >> Subject: memory requirements >> From: Ana.M.C.Beltran at ww.uni-erlangen.de >> Date: Wed, February 26, 2014 17:58 >> To: ilastik-user at ilastik.org >> -------------------------------------------------------------------------- >> >> Dear All, >> >> Is it a special memory recquired to run Ilastik 0.5.012? or is there any >> limit for the files to use? is there any settings I can modify to work >> quicker? >> I am trying to segment tomography data and for files around 300 MB the >> software doesn?t work, and the computer have a big memory. >> Could you give me some ideas? >> Thank you very much in advance >> best regards >> >> - >> Dr.Ing. Ana M Beltran >> >> >> >> >> > From burcin at erocal.org Wed Mar 5 23:15:27 2014 From: burcin at erocal.org (Burcin Erocal) Date: Wed, 5 Mar 2014 23:15:27 +0100 Subject: [ilastik-user] Google Summer of Code Projects - please advertise Message-ID: <20140305231527.50b5761c@carl.erocal.org> Hi, this year ilastik is participating in the Google Summer of Code [1] via the lmonade project [2]. [1] http://www.google-melange.com/gsoc/homepage/google/gsoc2014 [2] http://www.google-melange.com/gsoc/org2/google/gsoc2014/lmonade We have many exciting projects [3] for students to work (from home) over the summer, get mentored by experts in the field and get paid by Google [4]. [3] http://wiki.lmona.de/get_involved/projects [4] http://www.google-melange.com/gsoc/document/show/gsoc_program/google/gsoc2014/help_page#1._How_do_payments_work The application deadline for students is March 21st. These days they should be talking to potential mentors and forming their application. http://wiki.lmona.de/get_involved/gsoc#Prospective_Students If you know any student who might be interested or have access to any forum such students might read, please forward this announcement. Cheers, Burcin Possible project ideas related to ilastik: Software infrastructure projects - Use OpenGL for blending layers in Volumina - Multiscale views based on pyramids in volumina - Improve 2d/3d object browser in volumina - User Experience Improvements in ilastik - VIGRA Java bindings - Parallelization of the VIGRA library Algorithms & data structures projects - Lazy connected components in VIGRA - Image features for machine learning in VIGRA Detailed descriptions available here: http://wiki.lmona.de/get_involved/projects From christoph.sommer at imba.oeaw.ac.at Tue Mar 11 12:12:51 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Tue, 11 Mar 2014 12:12:51 +0100 Subject: [ilastik-user] [Fwd: memory requirements] In-Reply-To: <206f2bf765bbf5fe4c61702c777d45b9.squirrel@faumail.uni-erlangen.de> References: <5af6b9214448ff1717a15885f8b0f14a.squirrel@faumail.uni-erlangen.de> <53145F4C.50008@imba.oeaw.ac.at> <7ec5cfa8ebc94ac701bde807e3371445.squirrel@faumail.uni-erlangen.de> <5314A541.4080202@imba.oeaw.ac.at> <206f2bf765bbf5fe4c61702c777d45b9.squirrel@faumail.uni-erlangen.de> Message-ID: <531EEFB3.7050406@imba.oeaw.ac.at> Dear Ana, the features are not saved in the project file, just because they are too big. But the settings of which features you selected is saved. So, you just have to recompute them. Are you using the "Interactive Segmentation" module? Before I thought you are just using the Classification module. The Interactive Segmentation requires always user input (in form of markings for background and objects of interest). These should be in the project file (are they?). > I have to choose again the features Yes, there is currently no way around it > input the weights Segment This is for the "Interactive Segmentation", right? What kind of weights are you using? The Classification result? > mark any background. The markings are supposed to be saved... So you should be able to click "Segment" after you initialized the "Interactive Segmentation" with the weights... > Now, segment activates but the software doesn?t work...and start again. Is there an error message? Note, that you can always export results you obtained after working with ilastik. Just right-click on an Overlay (which you display) and click export. Cheers, Christoph > Dear Christoph > > I followed your advices and I minimize the data set and it runs. I have > one question. I saved the work as project and when I open it, I cannot > continue at the same point: the options are not activate. I have to > choose again the features, input the weights Segment, mark any background. > Now, segment activates but the software doesn?t work...and start again. > What I am doing wrong? > I spent many hours late week doing all the segmentation and now, I cannot > open it. > Thanks a lot > > best regards > > Ana > > > Christoph Sommer wrote: >> Hi Ana, >> >> that's a rather big dataset. E.g., The dimensions of (989,974,717) have >> 690,676,062 different voxels, which will use 659MB (8-bit) or 1.29GB >> (16-bit) in RAM. Each feature (used for classification) will produce >> extra memory requirement. E.g., A simple color feature at a certain >> scale will occupy 2.57GB in RAM (4 Byte per voxel)... >> >>> Is it better Windows 7? >> Hmm, I do not personally use XP, but I doubt that using to Win 7 will be >> game changer... >> >> I would recommend to split the data into sub-cubes of a few >> representative, say, 300^3 cubes for training the classifier and then >> apply this classifier to the full image in batch mode. >> >> Cheers, >> Christoph >> >> >>> Dear Christoph >>> >>> The images correspond to a STEM tomography serie. The pixel size is >>> (989,974,717). There is not temporal channel. >>> Z could be reduce to 550 pixels. >>> I have tried for (989,974,81) and it works slow but good. However, we >>> have >>> problems from higher values of Z and I need to work with the full data. >>> Is there any influence of the Operative System? We are using Windows XP. >>> Is it better Windows 7? >>> >>> Thank you very much for your help, >>> >>> Best regards >>> >>> Ana >>> >>> Christoph Sommer wrote: >>>> Dear Ana, >>>> >>>> what is the dimensionality of your data in pixels (x,y,z,c)? Is there >>>> also a temporal channel? >>>> >>>> ilastik 0.5 is still computing all outputs on the full image. For large >>>> images this can take very long, even if the image data fits into main >>>> memory. What people typically do, is to train the classifier on smaller >>>> subsets of the image or to downsample the data (if this is possible >>>> without loosing important information for the classification task at >>>> hand). Once a classifier is trained you can apply the it to bigger data >>>> in batch mode, which can run in the background (via Automate). >>>> >>>> Also note, that not only the image data itself has to fit into main >>>> memory, but also the image features and several intermediate results >>>> have to be allocated. Hence, using a large feature set, might lead to >>>> memory errors. >>>> >>>> Best, >>>> Christoph >>>> >>>> >>>>> Dear Christoph >>>>> >>>>> Today, I can report it says: Memory error, working with an original >>>>> file >>>>> of 184 MB It uses the 27GB of the computer. >>>>> Please, could you confirm me the memory requirements of the software? >>>>> I >>>>> am >>>>> very interesting on using it for my research. >>>>> >>>>> Thank you very much on advance >>>>> >>>>> Best regards >>>>> >>>>> Dr. Ing. Ana M Beltr?n >>>>> >>>>> >>>>> Ana.M.C.Beltran at ww.uni-erlangen.de wrote: >>>>>> Dear Christoph >>>>>> >>>>>> I can load the data. I load series of images beause I want to do 3D >>>>>> segmentation. When the total file is around 60 MB, it works slow but >>>>>> without problems. However, I cannot work with bigger files. >>>>>> I am using a computer with 24GB RAM, and the memory is completely >>>>>> saturated. It works in Windows 7, should I change to windows XP or >>>>>> another >>>>>> version? >>>>>> The file is a series of tiff images, 8 bits, gray scale >>>>>> >>>>>> Thank you very much for your support >>>>>> >>>>>> Best regards >>>>>> >>>>>> Ana M Beltr?n >>>>>> >>>>>> >>>>>> ---------------------------- Original Message >>>>>> ---------------------------- >>>>>> Subject: memory requirements >>>>>> From: Ana.M.C.Beltran at ww.uni-erlangen.de >>>>>> Date: Wed, February 26, 2014 17:58 >>>>>> To: ilastik-user at ilastik.org >>>>>> -------------------------------------------------------------------------- >>>>>> >>>>>> Dear All, >>>>>> >>>>>> Is it a special memory recquired to run Ilastik 0.5.012? or is there >>>>>> any >>>>>> limit for the files to use? is there any settings I can modify to >>>>>> work >>>>>> quicker? >>>>>> I am trying to segment tomography data and for files around 300 MB >>>>>> the >>>>>> software doesn?t work, and the computer have a big memory. >>>>>> Could you give me some ideas? >>>>>> Thank you very much in advance >>>>>> best regards >>>>>> >>>>>> - >>>>>> Dr.Ing. Ana M Beltran >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >> > From corrado.cali at kaust.edu.sa Tue May 20 14:23:49 2014 From: corrado.cali at kaust.edu.sa (Corrado Cali) Date: Tue, 20 May 2014 12:23:49 +0000 Subject: [ilastik-user] Ilastik export mesh error References: Message-ID: Hi all, I?ve been working a lot lately with the new version of iLastik (on a linux workstation), and I appreciate a lot the fact that I can segment independently each different objects and for each one I can define a separate background and define a different name. It comes very handy because I segment the entire neuropil from electron micrographs, and this is a major improvement. Nevertheless, more or less randomly whevener I export the obj file, iLastik returns an error and crashes, without exporting the file. I have joined the log file, the last 4 lines contains the error I got. Could you have a look at it and let me know if someone have experienced the same issue? Thanks a lot in advance, best regards. CC :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: Ing. Corrado Cal?, PhD Post Doctoral Fellow Laboratory of Cell Imaging and Energetics Division of BESE Room 4266, Level 4, Building 2 4700 King Abdullah University of Science and Technology Thuwal 23955-6900, Kingdom of Saudi Arabia Tel: +966 (0)2 808 2969 skype: menestrello82 Please note that week ends on the kingdom are on fridays and saturdays :~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~: Begin forwarded message: ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: ilastik_log.txt URL: From thorben.kroeger at iwr.uni-heidelberg.de Wed May 21 09:06:44 2014 From: thorben.kroeger at iwr.uni-heidelberg.de (Thorben Kroeger) Date: Wed, 21 May 2014 09:06:44 +0200 Subject: [ilastik-user] Ilastik export mesh error In-Reply-To: References: Message-ID: <537C5084.1030702@iwr.uni-heidelberg.de> Dear Corrado, On 20.05.2014 14:23, Corrado Cali wrote: > Nevertheless, more or less randomly whevener I export the obj file, > iLastik returns an error and crashes, without exporting the file. Thanks for trying the new ilastik, and thank you for the bug report. I have opened an issue here: https://github.com/ilastik/ilastik/issues/896 where you can track the progress. Cheers, Thorben From glendon.holst at kaust.edu.sa Tue Nov 11 12:24:32 2014 From: glendon.holst at kaust.edu.sa (Glendon Holst) Date: Tue, 11 Nov 2014 14:24:32 +0300 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <20141029131500.07ea9b92@carl.erocal.org> References: <544CCCB5.1060809@kaust.edu.sa> <20141029131500.07ea9b92@carl.erocal.org> Message-ID: <5461F1F0.7010502@kaust.edu.sa> Thanks Burcin! I tried your fix out and it worked. I appreciate that. I've been doing some testing with ilastik to try to work it into our pipeline. Enclosed is a PDF of the partial results (table and charts). The system specs for our testing were: 40 core Intel Xeon CPU E5-2680 v2 @ 2.80GHz 64 GB RAM Scientific Linux 6.5 For our purposes, the minimum data-set size is 26 GB. 4096X4096x1500 pixels, with ~1.75 pixels per cell boundary, covering ~100^3 microns (a function of minimum resolution, and capturing a full cell). For us, ilastik's Carving module can't handle these data sets directly. Extrapolating from its performance on < 2 GB stacks, processing the 26 GB stack would take ~1000 GB to pre-process, ~700 GB to segment, ~24 hours to pre-process, and ~15 minutes per interactive label segmentation. Memory usage is not the only issue with ilastik, much of its processing (on our builds) runs single-threaded. Small parts of the segmentation process do appear fully multi-threaded. Note: We need to build ilastik because SL (based on RHEL) is so old. There are a few small issues that occur, but we have workarounds which I'll pass along to the ilastik dev list later. We are looking for a tool like ilastik's Carving module but scalable to larger data-sets, running on HPC clusters. Our current workaround is to split the stacks into ~32 sub-stack chunks and manually integrate the segmentations from each chunk into TrakEM2, where they are managed and refined. While less than ideal, we still anticipate a significant speedup of our workflow. BTW, I do have a manual workflow between ilastik and TrakEM2, using ImageMagick and HDF5 Viewer to process the image sequences into masks, and to create the enumeration format used by ilastik for its fg_voxels and bg_voxels. I'll keep investigating ilastik in more depth later (e.g., importing labels, other data handling options, and multi-threading), when I'll likely have more questions. On 10/29/2014 03:15 PM, Burcin Erocal wrote: Hi Glendon, many thanks for the error report. I hope we can get ilastik to work for you eventually. :) I'm not familiar with the carving code or cylemon, but staring at the cylemon code I can guess what is going wrong. On Sun, 26 Oct 2014 13:28:05 +0300 Glendon Holst wrote: ... Looking at the print statements in the source [1], it looks like the error is generated around line 149, where it tries to create a numpy ndarray of size totalNeighborhoods. https://github.com/ilastik/cylemon/blob/master/cylemon/segmentation.pyx#L135 totalNeighborhoods is a C int, which might have taken negative values due to an overflow. Note that this is going to need a lot of RAM, since we want to initialize an array of size > 2^31, where each element needs 12 bytes (see line 147). I put up some changes in a fork of the cylemon repository to work around the overflow by extending totalNeighborhoods to be a C long. The code is in the branch named totalNeighborhoods_overflow: https://github.com/burcin/cylemon/tree/totalNeighborhoods_overflow Since you're working from a local build, perhaps you can try with this version and see if it makes it past the error. Please let us know how it goes or if you need any more help. Best, Burcin ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: Technical Paper - ilastik data.pdf Type: application/pdf Size: 86118 bytes Desc: not available URL: From anna.kreshuk at iwr.uni-heidelberg.de Wed Nov 12 16:27:38 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Wed, 12 Nov 2014 16:27:38 +0100 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <5461F1F0.7010502@kaust.edu.sa> References: <544CCCB5.1060809@kaust.edu.sa> <20141029131500.07ea9b92@carl.erocal.org> <5461F1F0.7010502@kaust.edu.sa> Message-ID: <54637C6A.5010200@iwr.uni-heidelberg.de> Hi Glendon, We have plans to shift the core of carving from its current cylemon implementation to the new vigra graph module in the coming months. We expect this will help with some of the performance issues (especially in what concerns preprocessing) and will in general make the code more maintainable. But it will, of course, not magically scale it to 26 GB datasets. As to running in a single thread, partially the fault here lies with the global nature of the watershed transform. However, we now have a blockwise implementation in vigra, which we plan to use for carving as well, but it isn't fully finished and polished yet. In summary, we are working on it, but there are no improvements we could push to you right now. Cheers, Anna On 11.11.2014 12:24, Glendon Holst wrote: > Thanks Burcin! > > I tried your fix out and it worked. I appreciate that. > > I've been doing some testing with ilastik to try to work it into our > pipeline. Enclosed is a PDF of the partial results (table and charts). > > The system specs for our testing were: > > 40 core Intel Xeon CPU E5-2680 v2 @ 2.80GHz > 64 GB RAM > Scientific Linux 6.5 > > > For our purposes, the minimum data-set size is 26 GB. 4096X4096x1500 > pixels, with ~1.75 pixels per cell boundary, covering ~100^3 microns > (a function of minimum resolution, and capturing a full cell). For > us, ilastik's Carving module can't handle these data sets directly. > Extrapolating from its performance on < 2 GB stacks, processing the 26 > GB stack would take ~1000 GB to pre-process, ~700 GB to segment, ~24 > hours to pre-process, and ~15 minutes per interactive label segmentation. > > Memory usage is not the only issue with ilastik, much of its > processing (on our builds) runs single-threaded. Small parts of the > segmentation process do appear fully multi-threaded. > > Note: We need to build ilastik because SL (based on RHEL) is so old. > There are a few small issues that occur, but we have workarounds which > I'll pass along to the ilastik dev list later. > > We are looking for a tool like ilastik's Carving module but scalable > to larger data-sets, running on HPC clusters. > > Our current workaround is to split the stacks into ~32 sub-stack > chunks and manually integrate the segmentations from each chunk into > TrakEM2, where they are managed and refined. While less than ideal, > we still anticipate a significant speedup of our workflow. > > BTW, I do have a manual workflow between ilastik and TrakEM2, using > ImageMagick and HDF5 Viewer to process the image sequences into masks, > and to create the enumeration format used by ilastik for its fg_voxels > and bg_voxels. > > I'll keep investigating ilastik in more depth later (e.g., importing > labels, other data handling options, and multi-threading), when I'll > likely have more questions. > > > > On 10/29/2014 03:15 PM, Burcin Erocal wrote: >> Hi Glendon, >> >> many thanks for the error report. I hope we can get ilastik to work for >> you eventually. :) >> >> I'm not familiar with the carving code or cylemon, but staring at the >> cylemon code I can guess what is going wrong. >> >> On Sun, 26 Oct 2014 13:28:05 +0300 >> Glendon Holst wrote: >> >>> ... >> Looking at the print statements in the source [1], it looks like the >> error is generated around line 149, where it tries to create a numpy >> ndarray of size totalNeighborhoods. >> >> https://github.com/ilastik/cylemon/blob/master/cylemon/segmentation.pyx#L135 >> >> totalNeighborhoods is a C int, which might have taken negative values >> due to an overflow. >> >> Note that this is going to need a lot of RAM, since we want to >> initialize an array of size > 2^31, where each element needs 12 bytes >> (see line 147). >> >> >> I put up some changes in a fork of the cylemon repository to work >> around the overflow by extending totalNeighborhoods to be a C long. >> The code is in the branch named totalNeighborhoods_overflow: >> >> https://github.com/burcin/cylemon/tree/totalNeighborhoods_overflow >> >> Since you're working from a local build, perhaps you can try with this >> version and see if it makes it past the error. >> >> >> Please let us know how it goes or if you need any more help. >> >> >> Best, >> Burcin > > > ------------------------------------------------------------------------ > > This message and its contents including attachments are intended > solely for the original recipient. If you are not the intended > recipient or have received this message in error, please notify me > immediately and delete this message from your computer system. Any > unauthorized use or distribution is prohibited. Please consider the > environment before printing this email. > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From fred.hamprecht at iwr.uni-heidelberg.de Wed Nov 12 19:28:37 2014 From: fred.hamprecht at iwr.uni-heidelberg.de (Fred Hamprecht) Date: Wed, 12 Nov 2014 20:28:37 +0200 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <54637C6A.5010200@iwr.uni-heidelberg.de> References: <544CCCB5.1060809@kaust.edu.sa> <20141029131500.07ea9b92@carl.erocal.org> <5461F1F0.7010502@kaust.edu.sa> <54637C6A.5010200@iwr.uni-heidelberg.de> Message-ID: <5463A6D5.5000301@iwr.uni-heidelberg.de> Hello Glendon, thank you for your careful analysis and sharing these results. On performance: even the single threaded, RAM limited implementation was around ten times faster in preprocessing in version 0.5 than in 0.6 (before the code became more "beautiful"). But as Anna said, even a factor of 10 won't be enough on your data. With 26GB, you certainly qualify as power user -- the most powerful we've seen so far :-) Best regards, Fred On 12-Nov-14 17:27, Anna Kreshuk wrote: > Hi Glendon, > > We have plans to shift the core of carving from its current cylemon > implementation to the new vigra graph module in the coming months. We > expect this will help with some of the performance issues (especially in > what concerns preprocessing) and will in general make the code more > maintainable. But it will, of course, not magically scale it to 26 GB > datasets. As to running in a single thread, partially the fault here > lies with the global nature of the watershed transform. However, we now > have a blockwise implementation in vigra, which we plan to use for > carving as well, but it isn't fully finished and polished yet. > > In summary, we are working on it, but there are no improvements we could > push to you right now. > > Cheers, > Anna > > On 11.11.2014 12:24, Glendon Holst wrote: >> Thanks Burcin! >> >> I tried your fix out and it worked. I appreciate that. >> >> I've been doing some testing with ilastik to try to work it into our >> pipeline. Enclosed is a PDF of the partial results (table and charts). >> >> The system specs for our testing were: >> >> 40 core Intel Xeon CPU E5-2680 v2 @ 2.80GHz >> 64 GB RAM >> Scientific Linux 6.5 >> >> >> For our purposes, the minimum data-set size is 26 GB. 4096X4096x1500 >> pixels, with ~1.75 pixels per cell boundary, covering ~100^3 microns >> (a function of minimum resolution, and capturing a full cell). For >> us, ilastik's Carving module can't handle these data sets directly. >> Extrapolating from its performance on < 2 GB stacks, processing the 26 >> GB stack would take ~1000 GB to pre-process, ~700 GB to segment, ~24 >> hours to pre-process, and ~15 minutes per interactive label segmentation. >> >> Memory usage is not the only issue with ilastik, much of its >> processing (on our builds) runs single-threaded. Small parts of the >> segmentation process do appear fully multi-threaded. >> >> Note: We need to build ilastik because SL (based on RHEL) is so old. >> There are a few small issues that occur, but we have workarounds which >> I'll pass along to the ilastik dev list later. >> >> We are looking for a tool like ilastik's Carving module but scalable >> to larger data-sets, running on HPC clusters. >> >> Our current workaround is to split the stacks into ~32 sub-stack >> chunks and manually integrate the segmentations from each chunk into >> TrakEM2, where they are managed and refined. While less than ideal, >> we still anticipate a significant speedup of our workflow. >> >> BTW, I do have a manual workflow between ilastik and TrakEM2, using >> ImageMagick and HDF5 Viewer to process the image sequences into masks, >> and to create the enumeration format used by ilastik for its fg_voxels >> and bg_voxels. >> >> I'll keep investigating ilastik in more depth later (e.g., importing >> labels, other data handling options, and multi-threading), when I'll >> likely have more questions. >> >> >> >> On 10/29/2014 03:15 PM, Burcin Erocal wrote: >>> Hi Glendon, >>> >>> many thanks for the error report. I hope we can get ilastik to work for >>> you eventually. :) >>> >>> I'm not familiar with the carving code or cylemon, but staring at the >>> cylemon code I can guess what is going wrong. >>> >>> On Sun, 26 Oct 2014 13:28:05 +0300 >>> Glendon Holst wrote: >>> >>>> ... >>> Looking at the print statements in the source [1], it looks like the >>> error is generated around line 149, where it tries to create a numpy >>> ndarray of size totalNeighborhoods. >>> >>> https://github.com/ilastik/cylemon/blob/master/cylemon/segmentation.pyx#L135 >>> >>> totalNeighborhoods is a C int, which might have taken negative values >>> due to an overflow. >>> >>> Note that this is going to need a lot of RAM, since we want to >>> initialize an array of size > 2^31, where each element needs 12 bytes >>> (see line 147). >>> >>> >>> I put up some changes in a fork of the cylemon repository to work >>> around the overflow by extending totalNeighborhoods to be a C long. >>> The code is in the branch named totalNeighborhoods_overflow: >>> >>> https://github.com/burcin/cylemon/tree/totalNeighborhoods_overflow >>> >>> Since you're working from a local build, perhaps you can try with this >>> version and see if it makes it past the error. >>> >>> >>> Please let us know how it goes or if you need any more help. >>> >>> >>> Best, >>> Burcin >> >> >> ------------------------------------------------------------------------ >> >> This message and its contents including attachments are intended >> solely for the original recipient. If you are not the intended >> recipient or have received this message in error, please notify me >> immediately and delete this message from your computer system. Any >> unauthorized use or distribution is prohibited. Please consider the >> environment before printing this email. >> >> >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From siggiae at mail.rockefeller.edu Thu Nov 13 00:01:54 2014 From: siggiae at mail.rockefeller.edu (Eric D. Siggia) Date: Wed, 12 Nov 2014 23:01:54 +0000 Subject: [ilastik-user] missing monkey_patches for command line batch processing? Message-ID: <3389A153-4D9A-4AF6-B559-F20F84A6A0D1@rockefeller.edu> dear sirs, sorry to bother multiple people for what must be an old question, but can not find the answer on anything I could search via google. I was following the instructions on http://ilastik.org/documentation/pixelclassification/headless.html and operating with the newest release of ilastik, 1.1.3 on a mac running 10.9.4. I have no problem running batch processing via the web interface. I just downloaded the ilastik package and moved to my /Applications folder. I imagine some paths are not set, but can find no trace of the monkey_patches module on my computer. eric siggia ERROR MESSAGE: Desktop> /Applications//ilastik.app/Contents/MacOS/ilastik --headless --project=testDcrop.ilp "D.tiff" Traceback (most recent call last): File "/Applications//ilastik.app/Contents/Resources/__boot__.py", line 43, in _run() File "/Applications//ilastik.app/Contents/Resources/__boot__.py", line 38, in _run exec(compile(source, path, 'exec'), globals(), globals()) File "/Applications//ilastik.app/Contents/Resources/ilastik.py", line 24, in import ilastik_main File "ilastik_main.pyc", line 4, in File "/Applications/ilastik.app/Contents/Resources/ilastik.py", line 27, in import ilastik.monkey_patches ImportError: No module named monkey_patches 2014-11-12 17:31:59.067 ilastik[749:303] ilastik Error Prof. Eric D. Siggia Dept of Physics Center for Studies in Physics and Biology 523 Clark Hall Box 25 Cornell University Rockefeller University Ithaca, NY 14853-2501 1230 York Avenue New York, N.Y. 10065-6399 607 255 4669 212 327 8546 FAX 607 255 6428 FAX 212 327 8544 http://www.physics.rockefeller.edu/siggia From anna.kreshuk at iwr.uni-heidelberg.de Wed Nov 19 14:07:15 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Wed, 19 Nov 2014 14:07:15 +0100 Subject: [ilastik-user] missing monkey_patches for command line batch processing? In-Reply-To: <3389A153-4D9A-4AF6-B559-F20F84A6A0D1@rockefeller.edu> References: <3389A153-4D9A-4AF6-B559-F20F84A6A0D1@rockefeller.edu> Message-ID: <546C9603.80807@iwr.uni-heidelberg.de> Dear Dr. Siggia, Thanks for the error report, we haven't seen this one yet. monkey_patches is an internal ilastik module, it seems like something went wrong in the binary generation process. We'll be generating new binaries for a minor release soon and we'll let you know as soon as it's available. Meanwhile, the previous version, 1.1.2, should still be available on our file server: http://files.ilastik.org/ilastik-1.1.2-OSX.zip, so please try using this one if you need it to run before the new binaries are made. Best regards, Anna Kreshuk On 13.11.2014 00:01, Eric D. Siggia wrote: > dear sirs, > sorry to bother multiple people for what must be an old question, but can not find the answer on anything I could search via google. > > I was following the instructions on http://ilastik.org/documentation/pixelclassification/headless.html and operating with the newest release of ilastik, 1.1.3 on a mac running 10.9.4. I have no problem running batch processing via the web interface. I just downloaded the ilastik package and moved to my /Applications folder. I imagine some paths are not set, but can find no trace of the monkey_patches module on my computer. > > eric siggia > > ERROR MESSAGE: > > Desktop> /Applications//ilastik.app/Contents/MacOS/ilastik --headless --project=testDcrop.ilp "D.tiff" > Traceback (most recent call last): > File "/Applications//ilastik.app/Contents/Resources/__boot__.py", line 43, in > _run() > File "/Applications//ilastik.app/Contents/Resources/__boot__.py", line 38, in _run > exec(compile(source, path, 'exec'), globals(), globals()) > File "/Applications//ilastik.app/Contents/Resources/ilastik.py", line 24, in > import ilastik_main > File "ilastik_main.pyc", line 4, in > File "/Applications/ilastik.app/Contents/Resources/ilastik.py", line 27, in > import ilastik.monkey_patches > ImportError: No module named monkey_patches > 2014-11-12 17:31:59.067 ilastik[749:303] ilastik Error > > > > > Prof. Eric D. Siggia > Dept of Physics Center for Studies in Physics and Biology > 523 Clark Hall Box 25 > Cornell University Rockefeller University > Ithaca, NY 14853-2501 1230 York Avenue > New York, N.Y. 10065-6399 > 607 255 4669 212 327 8546 > FAX 607 255 6428 FAX 212 327 8544 > http://www.physics.rockefeller.edu/siggia > > > > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From Cameron.Nowell at monash.edu Thu Oct 9 00:17:30 2014 From: Cameron.Nowell at monash.edu (Cameron Nowell) Date: Wed, 08 Oct 2014 22:17:30 +0000 Subject: [ilastik-user] Batch Processing many hundreds of files Message-ID: <69A36BE875B3674093D86BD7E716D57C19F21C1F@PRK-MBX-04.vcp.local> Hi List, I want to process a whole pile of files (in the end 1000s of them). Is there any way to classify them all without having to load them into Ilastik? If I try and put too many files in through the batch prediction part it all comes to a grinding halt 9after chewing up 10 gig of my RAM). Cheers Cam Cameron J. Nowell Research Facilities Manager Imaging, FACS and Analysis Core Monash Institute of Pharmaceutical Sciences Monash University 399 Royal Parade (Mail address: 381 Royal Parade) Parkville, VIC, 3052 Australia Email: cameron.nowell at monash.edu Mobile: +61 422882700 Office: +61 9903 9587 LinkedIn: Profile Research Gate: Profile [BANNER_foremail] -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.jpg Type: image/jpeg Size: 50741 bytes Desc: image001.jpg URL: From bergs at janelia.hhmi.org Thu Oct 9 01:34:54 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Wed, 8 Oct 2014 23:34:54 +0000 Subject: [ilastik-user] Batch Processing many hundreds of files Message-ID: An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.jpg Type: image/jpeg Size: 50741 bytes Desc: image001.jpg URL: From Cameron.Nowell at monash.edu Thu Oct 9 01:37:37 2014 From: Cameron.Nowell at monash.edu (Cameron Nowell) Date: Wed, 08 Oct 2014 23:37:37 +0000 Subject: [ilastik-user] Batch Processing many hundreds of files In-Reply-To: References: Message-ID: <69A36BE875B3674093D86BD7E716D57C19F21CE7@PRK-MBX-04.vcp.local> Hi Stuart, Ah ok, thanks. Will look into the headless mode scripting. Its got to be beter than manually feeding in batched of 100 images. Cheers Cam From: Berg, Stuart [mailto:bergs at janelia.hhmi.org] Sent: Thursday, 9 October 2014 10:35 AM To: Cameron Nowell; ilastik-user at ilastik.org Subject: Re: [ilastik-user] Batch Processing many hundreds of files Hi Cam, Unfortunately, there is an issue with the batch processing mechanism in the current version. We'll let you know when it's fixed. In the mean time, I can suggest a workaround. You might try writing a shell script to process your data using ilastik's headless mode, described here: http://ilastik.org/documentation/pixelclassification/headless.html The issue you are experiencing in the GUI also affects headless mode, but if you write a script to process your images in batches, e.g. 50 images at a time, then you might be able to get the results you need. Best regards, Stuart From: Cameron Nowell > Date: Wednesday, October 8, 2014 6:17 PM To: "ilastik-user at ilastik.org" > Subject: [ilastik-user] Batch Processing many hundreds of files Hi List, I want to process a whole pile of files (in the end 1000s of them). Is there any way to classify them all without having to load them into Ilastik? If I try and put too many files in through the batch prediction part it all comes to a grinding halt 9after chewing up 10 gig of my RAM). Cheers Cam Cameron J. Nowell Research Facilities Manager Imaging, FACS and Analysis Core Monash Institute of Pharmaceutical Sciences Monash University 399 Royal Parade (Mail address: 381 Royal Parade) Parkville, VIC, 3052 Australia Email: cameron.nowell at monash.edu Mobile: +61 422882700 Office: +61 9903 9587 LinkedIn: Profile Research Gate: Profile [BANNER_foremail] -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.jpg Type: image/jpeg Size: 50741 bytes Desc: image001.jpg URL: From glendon.holst at kaust.edu.sa Sun Oct 26 11:28:05 2014 From: glendon.holst at kaust.edu.sa (Glendon Holst) Date: Sun, 26 Oct 2014 13:28:05 +0300 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored Message-ID: <544CCCB5.1060809@kaust.edu.sa> Hi! I encountered the following crash while testing the ilastik Carving module on a 1024x1024x1528 8-bit grayscale stack. I was running a local build (v1.1.3 via Buildem) on Scientific Linux 6.5, on a machine with 256 GB memory, and 40 cores. I had already used this build successfully to carve on a 512x512x1528 stack (on a 40 core, 64 GB machine). I had already loaded the images and was performing the processing step (with black lines and feature size 1.75). It was an hour into the process when a thread/stack-dump was displayed. The essential bits were: Watershed...uint8 version initializing queues 100.0000% watersheds 100.0000% Constructing RAG... maximum label = 10541273 neighborCount: 40.211769 MB counting nhood sizes: 26.529467 sec. neighborOffset: 40.211769 MB offsetBackup: 40.211769 MB Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored Fatal Python error: Segmentation fault ... Thread 0x00007f064526c700: File "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik/shell/gui/startShellGui...", line 69 in startShellGui File "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik_main.py", line 86 in main File "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik.py", line 44 in main File "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik.py", line 50 in /var/remote/projects/epfl/software/ilastik/1.1.3/bin/ilastik_gui: line 7: 4033 Segmentation fault $BUILDEM_DIR/bin/python2.7 $ILASTIK_PATH/ilastik/ilastik.py "$@" This info did not appear in the .ilastik_log.txt file. This was the relevant portion of the log file: INFO 2014-10-26 09:46:58,470 ilastik_main 4033 139664906700544 Starting ilastik from "/var/remote/projects/epfl/software/ilastik/1.1.3/src". INFO 2014-10-26 09:47:01,731 opThresholdTwoLevels 4033 139664906700544 Using 'vigra' labeling implemetation WARNING 2014-10-26 09:47:02,157 __init__ 4033 139664906700544 Failed to import connected components workflow; check dependencies: No module named connectedComponents WARNING 2014-10-26 09:47:02,183 __init__ 4033 139664906700544 Failed to import automatic tracking workflow (conservation tracking). For this workflow, see the installationinstructions on our website ilastik.org; check dependencies: No module named pgmlink INFO 2014-10-26 09:47:12,785 projectManager 4033 139664906700544 Opening Project: /home/holstgr/Documents/Segmentation/KB-E0010/ilastik-stack-split-xy4-x0y0.ilp INFO 2014-10-26 09:48:48,179 opPreprocessing 4033 139664034821888 input volume shape: (1024, 1024, 1528) INFO 2014-10-26 09:48:48,180 opPreprocessing 4033 139664034821888 input volume size: (1528,) MB INFO 2014-10-26 09:48:50,949 opPreprocessing 4033 139664034821888 applying filter on shape = (1024, 1024, 1528) INFO 2014-10-26 09:48:50,950 opPreprocessing 4033 139664034821888 greatest eigenvalue of Hessian of Gaussian INFO 2014-10-26 09:49:03,624 arrayCacheMemoryMgr 4033 139663026452224 RAM: 51.312 GB (26%), CPU: Avg=06%, [9.2, 9.3, 0.0, 21.0, 1.0, 0.0, 1.0, 0.0, 95.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 4.0, 1.0, 16.0, 17.2, 0.0, 2.9, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] INFO 2014-10-26 10:04:51,307 arrayCacheMemoryMgr 4033 139663026452224 RAM: 69.219 GB (38%), CPU: Avg=03%, [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] INFO 2014-10-26 10:13:17,625 opPreprocessing 4033 139664034821888 Filter took 1466.673496 seconds INFO 2014-10-26 10:13:22,857 arrayCacheMemoryMgr 4033 139663026452224 RAM: 8.039 GB (03%), CPU: Avg=05%, [17.0, 16.8, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 92.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 8.0, 0.0, 0.0, 17.6, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] INFO 2014-10-26 10:13:45,361 arrayCacheMemoryMgr 4033 139663026452224 RAM: 45.166 GB (22%), CPU: Avg=03%, [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] INFO 2014-10-26 10:35:46,988 opPreprocessing 4033 139664034821888 done 10541273 INFO 2014-10-26 10:35:46,990 opPreprocessing 4033 139664034821888 Watershed took 1330.572929 seconds INFO 2014-10-26 10:35:47,776 arrayCacheMemoryMgr 4033 139663026452224 RAM: 25.946 GB (12%), CPU: Avg=03%, [0.0, 0.0, 0.0, 3.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 75.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 22.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] This didn't seem to be an out-of-memory issue (I've had those before :-) If you have any suggestions for how I can help track down the issue, I'd be happy to help. Thank you. Glendon ________________________________ This message and its contents including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email. From ullrich.koethe at iwr.uni-heidelberg.de Wed Oct 29 13:12:20 2014 From: ullrich.koethe at iwr.uni-heidelberg.de (Ullrich Koethe) Date: Wed, 29 Oct 2014 13:12:20 +0100 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <544CCCB5.1060809@kaust.edu.sa> References: <544CCCB5.1060809@kaust.edu.sa> Message-ID: <5450D9A4.6020109@iwr.uni-heidelberg.de> Hi Glendon, from the outside this looks like an integer overflow on an int32 variable which is not straightforward to fix. A general revision of the carving workflow is high on our priority list, but has not been scheduled concretely yet. If you want to try to track down the problem yourself, we will certainly happy to help. Best regards Ulli Am 26.10.2014 um 11:28 schrieb Glendon Holst: > Hi! > > I encountered the following crash while testing the ilastik Carving > module on a 1024x1024x1528 8-bit grayscale stack. I was running a local > build (v1.1.3 via Buildem) on Scientific Linux 6.5, on a machine with > 256 GB memory, and 40 cores. I had already used this build successfully > to carve on a 512x512x1528 stack (on a 40 core, 64 GB machine). > > I had already loaded the images and was performing the processing step > (with black lines and feature size 1.75). It was an hour into the > process when a thread/stack-dump was displayed. The essential bits were: > > Watershed...uint8 version > > initializing queues 100.0000% > > watersheds 100.0000% > > Constructing RAG... > > maximum label = 10541273 > > neighborCount: 40.211769 MB > > counting nhood sizes: 26.529467 sec. > > neighborOffset: 40.211769 MB > > offsetBackup: 40.211769 MB > > Exception ValueError: 'negative dimensions are not allowed' in > 'cylemon.segmentation.arcMapByLabels' ignored > > Fatal Python error: Segmentation fault > > ... > > Thread 0x00007f064526c700: > > File > "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik/shell/gui/startShellGui...", > line 69 in startShellGui > > File > "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik_main.py", > line 86 in main > > File > "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik.py", > line 44 in main > > File > "/var/remote/projects/epfl/software/ilastik/1.1.3/src/ilastik/ilastik/ilastik.py", > line 50 in > > /var/remote/projects/epfl/software/ilastik/1.1.3/bin/ilastik_gui: line > 7: 4033 Segmentation fault $BUILDEM_DIR/bin/python2.7 > $ILASTIK_PATH/ilastik/ilastik.py "$@" > > > This info did not appear in the .ilastik_log.txt file. This was the > relevant portion of the log file: > > INFO 2014-10-26 09:46:58,470 ilastik_main 4033 139664906700544 Starting > ilastik from "/var/remote/projects/epfl/software/ilastik/1.1.3/src". > > INFO 2014-10-26 09:47:01,731 opThresholdTwoLevels 4033 139664906700544 > Using 'vigra' labeling implemetation > > WARNING 2014-10-26 09:47:02,157 __init__ 4033 139664906700544 Failed to > import connected components workflow; check dependencies: No module > named connectedComponents > > WARNING 2014-10-26 09:47:02,183 __init__ 4033 139664906700544 Failed to > import automatic tracking workflow (conservation tracking). For this > workflow, see the installationinstructions on our website ilastik.org; > check dependencies: No module named pgmlink > > INFO 2014-10-26 09:47:12,785 projectManager 4033 139664906700544 Opening > Project: > /home/holstgr/Documents/Segmentation/KB-E0010/ilastik-stack-split-xy4-x0y0.ilp > > > INFO 2014-10-26 09:48:48,179 opPreprocessing 4033 139664034821888 input > volume shape: (1024, 1024, 1528) > > INFO 2014-10-26 09:48:48,180 opPreprocessing 4033 139664034821888 input > volume size: (1528,) MB > > INFO 2014-10-26 09:48:50,949 opPreprocessing 4033 139664034821888 > applying filter on shape = (1024, 1024, 1528) > > INFO 2014-10-26 09:48:50,950 opPreprocessing 4033 139664034821888 > greatest eigenvalue of Hessian of Gaussian > > INFO 2014-10-26 09:49:03,624 arrayCacheMemoryMgr 4033 139663026452224 > RAM: 51.312 GB (26%), CPU: Avg=06%, [9.2, 9.3, 0.0, 21.0, 1.0, 0.0, 1.0, > 0.0, 95.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 4.0, 1.0, 16.0, 17.2, 0.0, > 2.9, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] > > INFO 2014-10-26 10:04:51,307 arrayCacheMemoryMgr 4033 139663026452224 > RAM: 69.219 GB (38%), CPU: Avg=03%, [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] > > INFO 2014-10-26 10:13:17,625 opPreprocessing 4033 139664034821888 Filter > took 1466.673496 seconds > > INFO 2014-10-26 10:13:22,857 arrayCacheMemoryMgr 4033 139663026452224 > RAM: 8.039 GB (03%), CPU: Avg=05%, [17.0, 16.8, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 92.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 8.0, 0.0, > 0.0, 17.6, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] > > INFO 2014-10-26 10:13:45,361 arrayCacheMemoryMgr 4033 139663026452224 > RAM: 45.166 GB (22%), CPU: Avg=03%, [1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 100.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] > > INFO 2014-10-26 10:35:46,988 opPreprocessing 4033 139664034821888 done > 10541273 > > INFO 2014-10-26 10:35:46,990 opPreprocessing 4033 139664034821888 > Watershed took 1330.572929 seconds > > INFO 2014-10-26 10:35:47,776 arrayCacheMemoryMgr 4033 139663026452224 > RAM: 25.946 GB (12%), CPU: Avg=03%, [0.0, 0.0, 0.0, 3.0, 0.0, 0.0, 0.0, > 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 75.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, > 0.0, 0.0, 0.0, 22.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] > > > This didn't seem to be an out-of-memory issue (I've had those before > :-) If you have any suggestions for how I can help track down the > issue, I'd be happy to help. > > Thank you. > > Glendon > > > > ________________________________ > > This message and its contents including attachments are intended solely > for the original recipient. If you are not the intended recipient or > have received this message in error, please notify me immediately and > delete this message from your computer system. Any unauthorized use or > distribution is prohibited. Please consider the environment before > printing this email. > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user From burcin at erocal.org Wed Oct 29 13:15:00 2014 From: burcin at erocal.org (Burcin Erocal) Date: Wed, 29 Oct 2014 13:15:00 +0100 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <544CCCB5.1060809@kaust.edu.sa> References: <544CCCB5.1060809@kaust.edu.sa> Message-ID: <20141029131500.07ea9b92@carl.erocal.org> Hi Glendon, many thanks for the error report. I hope we can get ilastik to work for you eventually. :) I'm not familiar with the carving code or cylemon, but staring at the cylemon code I can guess what is going wrong. On Sun, 26 Oct 2014 13:28:05 +0300 Glendon Holst wrote: > I encountered the following crash while testing the ilastik Carving > module on a 1024x1024x1528 8-bit grayscale stack. I was running a > local build (v1.1.3 via Buildem) on Scientific Linux 6.5, on a > machine with 256 GB memory, and 40 cores. I had already used this > build successfully to carve on a 512x512x1528 stack (on a 40 core, 64 > GB machine). > > I had already loaded the images and was performing the processing step > (with black lines and feature size 1.75). It was an hour into the > process when a thread/stack-dump was displayed. The essential bits > were: > > Watershed...uint8 version > > initializing queues 100.0000% > > watersheds 100.0000% > > Constructing RAG... > > maximum label = 10541273 > > neighborCount: 40.211769 MB > > counting nhood sizes: 26.529467 sec. > > neighborOffset: 40.211769 MB > > offsetBackup: 40.211769 MB > > Exception ValueError: 'negative dimensions are not allowed' in > 'cylemon.segmentation.arcMapByLabels' ignored > > Fatal Python error: Segmentation fault Looking at the print statements in the source [1], it looks like the error is generated around line 149, where it tries to create a numpy ndarray of size totalNeighborhoods. https://github.com/ilastik/cylemon/blob/master/cylemon/segmentation.pyx#L135 totalNeighborhoods is a C int, which might have taken negative values due to an overflow. Note that this is going to need a lot of RAM, since we want to initialize an array of size > 2^31, where each element needs 12 bytes (see line 147). I put up some changes in a fork of the cylemon repository to work around the overflow by extending totalNeighborhoods to be a C long. The code is in the branch named totalNeighborhoods_overflow: https://github.com/burcin/cylemon/tree/totalNeighborhoods_overflow Since you're working from a local build, perhaps you can try with this version and see if it makes it past the error. Please let us know how it goes or if you need any more help. Best, Burcin From ullrich.koethe at iwr.uni-heidelberg.de Wed Oct 29 13:30:48 2014 From: ullrich.koethe at iwr.uni-heidelberg.de (Ullrich Koethe) Date: Wed, 29 Oct 2014 13:30:48 +0100 Subject: [ilastik-user] Exception ValueError: 'negative dimensions are not allowed' in 'cylemon.segmentation.arcMapByLabels' ignored In-Reply-To: <20141029131500.07ea9b92@carl.erocal.org> References: <544CCCB5.1060809@kaust.edu.sa> <20141029131500.07ea9b92@carl.erocal.org> Message-ID: <5450DDF8.4020103@iwr.uni-heidelberg.de> Hi Burcin, thanks for taking care of this problem. Just one remark: > I put up some changes in a fork of the cylemon repository to work > around the overflow by extending totalNeighborhoods to be a C long. I'd prefer you to use an explicitly sized int (i.e. 'int64_t' or whatever portable definition is available). Best Ulli From a3perez at ucsd.edu Thu Sep 4 20:30:00 2014 From: a3perez at ucsd.edu (Alex Perez) Date: Thu, 4 Sep 2014 11:30:00 -0700 Subject: [ilastik-user] Importing labels for pixel classification Message-ID: Dear All, I have a stack of binary label images that I would like to import as labels into Ilastik's Pixel Classification workflow. Is this possible? I am using Ilastik version 1.1.1. Thank you, Alex -------------- next part -------------- An HTML attachment was scrubbed... URL: From szymon.stoma at gmail.com Mon Sep 8 18:32:21 2014 From: szymon.stoma at gmail.com (Szymon Stoma) Date: Mon, 8 Sep 2014 18:32:21 +0200 Subject: [ilastik-user] Questions related to RGB image segmentation Message-ID: Dear developers, what is the status of using multichannel tifs in your software? I would like to know the recommendation how to work with ilastik to segment multidimensional images. Currently I tried xyc .tiff file and c is interpreted as z, and I do not know where / if I can change it. I converted the file to .png and it worked as I wanted. I also tried converting to RGB - it also worked. However I observe certain subtleties in behavior. Since I do not understand it very well, could you address the following questions: 1. What format do you recommend for xyc microscopy? 2. What format do you recommend for xycz microscopy? How I can operate on channels (change from z to c; if it is not possible how I should set it up e.g. in ImageJ). PS Thank you for great software! It is really good job! PPS I am referring to the email (it is a kind of follow-up): """ Dear James, We are aware of this problem and we are working on fixing it. In general, if you have an RGB image loaded as RGB, that is, when ilastik does not mistake it for a z-stack, there is an option in the Input dialog to display the channels individually (screenshot attached, select "Grayscale" from the pointed drop-down menu). Using more than 3 channels is also no problem, as long as ilastik does not interpret them as a z-stack. """ __ Szymon STOMA INRIA Paris | ETH Zurich Lifeware | ScopeM T: +41766390449 W: http://stoma.name/ -------------- next part -------------- An HTML attachment was scrubbed... URL: From xavier.vasques at epfl.ch Wed Sep 10 16:17:07 2014 From: xavier.vasques at epfl.ch (Vasques Xavier) Date: Wed, 10 Sep 2014 14:17:07 +0000 Subject: [ilastik-user] Loading Images Message-ID: <18BB5954-7CCF-4CC9-9F35-7620818CCA79@epfl.ch> Hi, We are running Ilastik on a cluster and we have to performed a pixel classification on 2TB of data. In the folder we have a set of images (.tif) and when starting to load the images, the progress bar is not progressing at all (the size of the project file is increasing at the sometime). I tried different set of images, 1000 images, 500, 200 and it is still difficult to load (each image size is 8.3 MB). Do you have any idea how to deal with large set of images ? May I have to convert them ? Thanks a lot for your help, Xavier From amjoao at yahoo.com.br Fri Sep 12 13:22:59 2014 From: amjoao at yahoo.com.br (Joao Medeiros) Date: Fri, 12 Sep 2014 04:22:59 -0700 Subject: [ilastik-user] from ilastik 0.5 to 1.1 Message-ID: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> I'm a user from Ilastik 0.5 and now I'm trying to use the new version, but the things are very different. For example, I was able to load a stack from tiff images and I'm trying to segment the image, but I can' t find how to. Please someone have a walk through to segment a image from a stack? Kind regards, Joao Medeiros -------------- next part -------------- An HTML attachment was scrubbed... URL: From amjoao at yahoo.com.br Fri Sep 12 14:18:44 2014 From: amjoao at yahoo.com.br (Joao Medeiros) Date: Fri, 12 Sep 2014 05:18:44 -0700 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> Message-ID: <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> I will try to be more specific: I have a 3D volume and want to segment an object inside it (there are only 1 object). I'm trying to use "Pixel Classification + Object Classification", so I was able to: 1) Input Data 2) Feature selection 3) Training 4) Thresholding 5) Object Feature Selection but in Object Classification I only can add the labels (on left panel) but can't select on the images (as I did in the training), how to advance from this point? In version 0.5 from training I can get the segmentation, how to have something like that in 1.1 version? Kind regards, Joao Medeiros Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros escreveu: I'm a user from Ilastik 0.5 and now I'm trying to use the new version, but the things are very different. For example, I was able to load a stack from tiff images and I'm trying to segment the image, but I can' t find how to. Please someone have a walk through to segment a image from a stack? Kind regards, Joao Medeiros _______________________________________________ ilastik-user mailing list ilastik-user at ilastik.org http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Fri Sep 12 16:38:10 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Fri, 12 Sep 2014 16:38:10 +0200 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: References: Message-ID: <54130552.2000702@iwr.uni-heidelberg.de> Dear Alex, It is not possible to load your own labels interactively, but you can train like that in headless mode. Would you be interested in such non-interactive training? I could send you a script for it then. Best, Anna On 04.09.2014 20:30, Alex Perez wrote: > Dear All, > > I have a stack of binary label images that I would like to import as > labels into Ilastik's Pixel Classification workflow. Is this possible? > > I am using Ilastik version 1.1.1. > > Thank you, > Alex > > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Fri Sep 12 16:40:56 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Fri, 12 Sep 2014 16:40:56 +0200 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> Message-ID: <541305F8.6020206@iwr.uni-heidelberg.de> Hi Joao, If you only need to segment 1 object, you don't need object classification at all. In the Pixel Classification applet, there is already a layer for Segmentation of all labels (lower left corner). This one just performs a simple segmentation, taking the most probable label. If you have a different threshold in mind, or want to use hysteresis thresholding for your data, you can do the thresholding step, and then right-click on the "Final Output" layer and export it. Let me know if that's what you had in mind, I'll try to walk you through it if it still doesn't work. Best, Anna On 12.09.2014 14:18, Joao Medeiros wrote: > I will try to be more specific: I have a 3D volume and want to segment > an object inside it (there are only 1 object). > > I'm trying to use "Pixel Classification + Object Classification", so I > was able to: > > 1) Input Data > 2) Feature selection > 3) Training > 4) Thresholding > 5) Object Feature Selection > > but in Object Classification I only can add the labels (on left panel) > but can't select on the images (as I did in the training), how to > advance from this point? > > In version 0.5 from training I can get the segmentation, how to have > something like that in 1.1 version? > > Kind regards, > Joao Medeiros > > > > Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros > escreveu: > > > I'm a user from Ilastik 0.5 and now I'm trying to use the new version, > but the things are very different. > > For example, I was able to load a stack from tiff images and I'm > trying to segment the image, but I can' t find how to. > Please someone have a walk through to segment a image from a stack? > > Kind regards, > Joao Medeiros > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From amjoao at yahoo.com.br Fri Sep 12 17:42:47 2014 From: amjoao at yahoo.com.br (Joao Medeiros) Date: Fri, 12 Sep 2014 08:42:47 -0700 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <541305F8.6020206@iwr.uni-heidelberg.de> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> <541305F8.6020206@iwr.uni-heidelberg.de> Message-ID: <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> Thank you Anna, I've trained the data. How to get a preview of the segmented object? In the training screen there are four panels (3 for slices and 1 for the 3D visualization, as in the 0.5 version) how to view the segmented object there? Kind regards, Joao Em Sexta-feira, 12 de Setembro de 2014 15:41, Anna Kreshuk escreveu: Hi Joao, If you only need to segment 1 object, you don't need object classification at all. In the Pixel Classification applet, there is already a layer for Segmentation of all labels (lower left corner). This one just performs a simple segmentation, taking the most probable label. If you have a different threshold in mind, or want to use hysteresis thresholding for your data, you can do the thresholding step, and then right-click on the "Final Output" layer and export it. Let me know if that's what you had in mind, I'll try to walk you through it if it still doesn't work. Best, Anna On 12.09.2014 14:18, Joao Medeiros wrote: I will try to be more specific: I have a 3D volume and want to segment an object inside it (there are only 1 object). > > >I'm trying to use "Pixel Classification + Object Classification", so I was able to: > > >1) Input Data >2) Feature selection >3) Training >4) Thresholding >5) Object Feature Selection > > >but in Object Classification I only can add the labels (on left panel) but can't select on the images (as I did in the training), how to advance from this point? > > >In version 0.5 from training I can get the segmentation, how to have something like that in 1.1 version? > > >Kind regards, >Joao Medeiros > > > > > >Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros escreveu: > > > >I'm a user from Ilastik 0.5 and now I'm trying to use the new version, but the things are very different. > > >For example, I was able to load a stack from tiff images and I'm trying to segment the image, but I can' t find how to. >Please someone have a walk through to segment a image from a stack? > > >Kind regards, >Joao Medeiros >_______________________________________________ >ilastik-user mailing list >ilastik-user at ilastik.org >http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > > > >_______________________________________________ ilastik-user mailing list ilastik-user at ilastik.org http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: From a3perez at ucsd.edu Fri Sep 12 18:51:31 2014 From: a3perez at ucsd.edu (Alex Perez) Date: Fri, 12 Sep 2014 09:51:31 -0700 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: <54130552.2000702@iwr.uni-heidelberg.de> References: <54130552.2000702@iwr.uni-heidelberg.de> Message-ID: Dear Anna, Yes, I am interested in trying non-interactive training in that way. That would be great. Thanks! Best, Alex On Sep 12, 2014 7:38 AM, "Anna Kreshuk" wrote: > Dear Alex, > > It is not possible to load your own labels interactively, but you can > train like that in headless mode. Would you be interested in such > non-interactive training? I could send you a script for it then. > > Best, > Anna > > On 04.09.2014 20:30, Alex Perez wrote: > > Dear All, > > I have a stack of binary label images that I would like to import as > labels into Ilastik's Pixel Classification workflow. Is this possible? > > I am using Ilastik version 1.1.1. > > Thank you, > Alex > > > > > _______________________________________________ > ilastik-user mailing listilastik-user at ilastik.orghttp://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Sat Sep 13 10:39:31 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sat, 13 Sep 2014 10:39:31 +0200 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> <541305F8.6020206@iwr.uni-heidelberg.de> <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> Message-ID: <20140913103931.gsrxaj4u0ws08kcg@wwwmail.urz.uni-heidelberg.de> Dear Joao, I've made a screenshot of the Training applet, which shows where to click to see the segmentation. Either activate all segmentations at the same time by checking the upper checkbox, or you can do it label by label, by clicking on the little eye icons. If it shows just one uniform label color, it means you haven't trained the classifier yet, so just press "live update". Let me know if it doesn't work for you or if you have any other problems transitioning to the new version. Cheers, Anna Quoting Joao Medeiros : > Thank you Anna, > > I've trained the data. How to get a preview of the segmented object? > In the training screen there are four panels (3 for slices and 1 > for the 3D visualization, as in the 0.5 version) how to view the > segmented object there? > > Kind regards, > Joao > > > > Em Sexta-feira, 12 de Setembro de 2014 15:41, Anna Kreshuk > escreveu: > > > > Hi Joao, > > If you only need to segment 1 object, you don't need object > classification at all. In the Pixel Classification applet, there > is already a layer for Segmentation of all labels (lower left > corner). This one just performs a simple segmentation, taking the > most probable label. If you have a different threshold in mind, or > want to use hysteresis thresholding for your data, you can do the > thresholding step, and then right-click on the "Final Output" > layer and export it. > > Let me know if that's what you had in mind, I'll try to walk you > through it if it still doesn't work. > > Best, > Anna > > > On 12.09.2014 14:18, Joao Medeiros wrote: > > I will try to be more specific: I have a 3D volume and want to > segment an object inside it (there are only 1 object). >> >> >> I'm trying to use "Pixel Classification + Object Classification", >> so I was able to: >> >> >> 1) Input Data >> 2) Feature selection >> 3) Training >> 4) Thresholding >> 5) Object Feature Selection >> >> >> but in Object Classification I only can add the labels (on left >> panel) but can't select on the images (as I did in the training), >> how to advance from this point? >> >> >> In version 0.5 from training I can get the segmentation, how to >> have something like that in 1.1 version? >> >> >> Kind regards, >> Joao Medeiros >> >> >> >> >> >> Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros >> escreveu: >> >> >> >> I'm a user from Ilastik 0.5 and now I'm trying to use the new >> version, but the things are very different. >> >> >> For example, I was able to load a stack from tiff images and I'm >> trying to segment the image, but I can' t find how to. >> Please someone have a walk through to segment a image from a stack? >> >> >> Kind regards, >> Joao Medeiros >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >> >> >> >> >> _______________________________________________ > ilastik-user mailing list ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- A non-text attachment was scrubbed... Name: segmentation.png Type: image/png Size: 193594 bytes Desc: not available URL: From anna.kreshuk at iwr.uni-heidelberg.de Sat Sep 13 10:43:43 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sat, 13 Sep 2014 10:43:43 +0200 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: References: <54130552.2000702@iwr.uni-heidelberg.de> Message-ID: <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> Dear Alex, Could you check if there is a script, named "train_headless.py" in the bin subdirectory of your ilastik installation? If not, you'd have to wait a couple days more and we'll generate all the new binaries with it. If yes, just try to use it and write back to us if it doesn't work or doesn't cover your use case. Cheers, Anna Quoting Alex Perez : > Dear Anna, > > Yes, I am interested in trying non-interactive training in that way. That > would be great. Thanks! > > Best, > Alex > On Sep 12, 2014 7:38 AM, "Anna Kreshuk" > wrote: > >> Dear Alex, >> >> It is not possible to load your own labels interactively, but you can >> train like that in headless mode. Would you be interested in such >> non-interactive training? I could send you a script for it then. >> >> Best, >> Anna >> >> On 04.09.2014 20:30, Alex Perez wrote: >> >> Dear All, >> >> I have a stack of binary label images that I would like to import as >> labels into Ilastik's Pixel Classification workflow. Is this possible? >> >> I am using Ilastik version 1.1.1. >> >> Thank you, >> Alex >> >> >> >> >> _______________________________________________ >> ilastik-user mailing >> listilastik-user at ilastik.orghttp://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >> >> > From anna.kreshuk at iwr.uni-heidelberg.de Sat Sep 13 10:59:54 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sat, 13 Sep 2014 10:59:54 +0200 Subject: [ilastik-user] Questions related to RGB image segmentation In-Reply-To: References: Message-ID: <20140913105954.ld8kwcruo004w8ow@wwwmail.urz.uni-heidelberg.de> Dear Szymon, The error you refer to has been fixed already, if it still doesn't allow you to change, let us know, we probably missed something. To change the axis interpretation, double-click on the dataset in the "Data Input" applet, there should be a way to edit them there. That being said, I am surprised that you observe differences between the modes you tried. Is your data originally 3-channel? Do I get it right, that you have different results, when you process it as an RGB tiff file and a png file? With the same labels? For xyc data, the format shouldn't matter as long as the compression is lossless. For xyzc data (and for any dataset of substantial size) we really recommend converting to hdf5, as it allows chunking and should make your interactive processing much smoother. You should be able to convert by using our "Data pre-processing" workflow. Finally, if you don't see any of the options I mentioned above, you'll have to update to the newest ilastik version. It should all be in the current one on the website, but we'll have an even newer and better version early next week. Let me know if this answers your questions or if you have any new ones Anna Quoting Szymon Stoma : > Dear developers, > > what is the status of using multichannel tifs in your software? I would > like to know the recommendation how to work with ilastik to segment > multidimensional images. Currently I tried xyc .tiff file and c is > interpreted as z, and I do not know where / if I can change it. I converted > the file to .png and it worked as I wanted. I also tried converting to RGB > - it also worked. However I observe certain subtleties in behavior. Since > I do not understand it very well, could you address the following questions: > > 1. What format do you recommend for xyc microscopy? > > 2. What format do you recommend for xycz microscopy? How I can operate on > channels (change from z to c; if it is not possible how I should set it up > e.g. in ImageJ). > > > PS Thank you for great software! It is really good job! > > PPS I am referring to the email (it is a kind of follow-up): > > """ > Dear James, > > We are aware of this problem and we are working on fixing it. In > general, if you have an RGB image loaded as RGB, that is, when ilastik > does not mistake it for a z-stack, there is an option in the Input > dialog to display the channels individually (screenshot attached, select > "Grayscale" from the pointed drop-down menu). Using more than 3 channels > is also no problem, as long as ilastik does not interpret them as a > z-stack. > """ > > __ > Szymon STOMA > INRIA Paris | ETH Zurich > > Lifeware | ScopeM > > T: +41766390449 W: http://stoma.name/ > From a3perez at ucsd.edu Sun Sep 14 00:46:29 2014 From: a3perez at ucsd.edu (Alex Perez) Date: Sat, 13 Sep 2014 15:46:29 -0700 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> References: <54130552.2000702@iwr.uni-heidelberg.de> <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> Message-ID: Dear Anna, The "train_headless.py" script is not in the bin directory of my installation. I installed version 1.1.1 for Linux 64-bit. Thanks for your help with the new binaries. Thanks, Alex On Sat, Sep 13, 2014 at 1:43 AM, wrote: > Dear Alex, > > Could you check if there is a script, named "train_headless.py" in the bin > subdirectory of your ilastik installation? If not, you'd have to wait a > couple days more and we'll generate all the new binaries with it. If yes, > just try to use it and write back to us if it doesn't work or doesn't cover > your use case. > > Cheers, > Anna > > > Quoting Alex Perez : > > Dear Anna, >> >> Yes, I am interested in trying non-interactive training in that way. That >> would be great. Thanks! >> >> Best, >> Alex >> On Sep 12, 2014 7:38 AM, "Anna Kreshuk" > heidelberg.de> >> wrote: >> >> Dear Alex, >>> >>> It is not possible to load your own labels interactively, but you can >>> train like that in headless mode. Would you be interested in such >>> non-interactive training? I could send you a script for it then. >>> >>> Best, >>> Anna >>> >>> On 04.09.2014 20:30, Alex Perez wrote: >>> >>> Dear All, >>> >>> I have a stack of binary label images that I would like to import as >>> labels into Ilastik's Pixel Classification workflow. Is this possible? >>> >>> I am using Ilastik version 1.1.1. >>> >>> Thank you, >>> Alex >>> >>> >>> >>> >>> _______________________________________________ >>> ilastik-user mailing listilastik-user at ilastik.orghttp:// >>> mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >>> >>> >>> >>> >> > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From amjoao at yahoo.com.br Mon Sep 15 10:30:44 2014 From: amjoao at yahoo.com.br (Joao Medeiros) Date: Mon, 15 Sep 2014 01:30:44 -0700 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <20140913103931.gsrxaj4u0ws08kcg@wwwmail.urz.uni-heidelberg.de> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> <541305F8.6020206@iwr.uni-heidelberg.de> <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> <20140913103931.gsrxaj4u0ws08kcg@wwwmail.urz.uni-heidelberg.de> Message-ID: <1410769844.45625.YahooMailNeo@web164905.mail.bf1.yahoo.com> Hi Anna, Thanks for your help and I'm sorry for so many basic questions, but it isn't working yet. I've attached a screenshot showing the question. Kind regards, Joao Em S?bado, 13 de Setembro de 2014 9:39, "anna.kreshuk at iwr.uni-heidelberg.de" escreveu: Dear Joao, I've made a screenshot of the Training applet, which shows where to click to see the segmentation. Either activate all segmentations at the same time by checking the upper checkbox, or you can do it label by label, by clicking on the little eye icons. If it shows just one uniform label color, it means you haven't trained the classifier yet, so just press "live update". Let me know if it doesn't work for you or if you have any other problems transitioning to the new version. Cheers, Anna Quoting Joao Medeiros : > Thank you Anna, > > I've trained the data. How to get a preview of the segmented object? > In the training screen there are four panels (3 for slices and 1 > for the 3D visualization, as in the 0.5 version) how to view the > segmented object there? > > Kind regards, > Joao > > > > Em Sexta-feira, 12 de Setembro de 2014 15:41, Anna Kreshuk > escreveu: > > > > Hi Joao, > > If you only need to segment 1 object, you don't need object > classification at all. In the Pixel Classification applet, there > is already a layer for Segmentation of all labels (lower left > corner). This one just performs a simple segmentation, taking the > most probable label. If you have a different threshold in mind, or > want to use hysteresis thresholding for your data, you can do the > thresholding step, and then right-click on the "Final Output" > layer and export it. > > Let me know if that's what you had in mind, I'll try to walk you > through it if it still doesn't work. > > Best, > Anna > > > On 12.09.2014 14:18, Joao Medeiros wrote: > > I will try to be more specific: I have a 3D volume and want to > segment an object inside it (there are only 1 object). >> >> >> I'm trying to use "Pixel Classification + Object Classification", >> so I was able to: >> >> >> 1) Input Data >> 2) Feature selection >> 3) Training >> 4) Thresholding >> 5) Object Feature Selection >> >> >> but in Object Classification I only can add the labels (on left >> panel) but can't select on the images (as I did in the training), >> how to advance from this point? >> >> >> In version 0.5 from training I can get the segmentation, how to >> have something like that in 1.1 version? >> >> >> Kind regards, >> Joao Medeiros >> >> >> >> >> >> Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros >> escreveu: >> >> >> >> I'm a user from Ilastik 0.5 and now I'm trying to use the new >> version, but the things are very different. >> >> >> For example, I was able to load a stack from tiff images and I'm >> trying to segment the image, but I can' t find how to. >> Please someone have a walk through to segment a image from a stack? >> >> >> Kind regards, >> Joao Medeiros >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >> >> >> >> >> _______________________________________________ > ilastik-user mailing list ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: ilastik.png Type: image/png Size: 177043 bytes Desc: not available URL: From bergs at janelia.hhmi.org Mon Sep 15 15:18:16 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Mon, 15 Sep 2014 13:18:16 +0000 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: References: <54130552.2000702@iwr.uni-heidelberg.de> <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> Message-ID: An HTML attachment was scrubbed... URL: From anna.kreshuk at iwr.uni-heidelberg.de Mon Sep 15 16:10:05 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (Anna Kreshuk) Date: Mon, 15 Sep 2014 16:10:05 +0200 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <1410769844.45625.YahooMailNeo@web164905.mail.bf1.yahoo.com> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> <541305F8.6020206@iwr.uni-heidelberg.de> <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> <20140913103931.gsrxaj4u0ws08kcg@wwwmail.urz.uni-heidelberg.de> <1410769844.45625.YahooMailNeo@web164905.mail.bf1.yahoo.com> Message-ID: <5416F33D.8070202@iwr.uni-heidelberg.de> Hi Joao, Ah, now I get it! You are absolutely right, the little 3D display is not working in the new version. The reason is that we now support "infinite" datasets and we can't make an infinite (or even very large) 3D object. We could restrict it to the field of view, like we do with the probability map, but it's not done yet. The only workflow, where this feature is enabled is Carving, which assumes that data fits in RAM. I hope ilastik can still be useful for you without this feature. Best, Anna On 15.09.2014 10:30, Joao Medeiros wrote: > Hi Anna, > > Thanks for your help and I'm sorry for so many basic questions, but it > isn't working yet. I've attached a screenshot showing the question. > > Kind regards, > Joao > > > > > Em S?bado, 13 de Setembro de 2014 9:39, > "anna.kreshuk at iwr.uni-heidelberg.de" > escreveu: > > > Dear Joao, > > I've made a screenshot of the Training applet, which shows where to > click to see the segmentation. Either activate all segmentations at > the same time by checking the upper checkbox, or you can do it label > by label, by clicking on the little eye icons. If it shows just one > uniform label color, it means you haven't trained the classifier yet, > so just press "live update". > > Let me know if it doesn't work for you or if you have any other > problems transitioning to the new version. > > Cheers, > Anna > > Quoting Joao Medeiros >: > > > Thank you Anna, > > > > I've trained the data. How to get a preview of the segmented object? > > In the training screen there are four panels (3 for slices and 1 > > for the 3D visualization, as in the 0.5 version) how to view the > > segmented object there? > > > > Kind regards, > > Joao > > > > > > > > Em Sexta-feira, 12 de Setembro de 2014 15:41, Anna Kreshuk > > > escreveu: > > > > > > > > Hi Joao, > > > > If you only need to segment 1 object, you don't need object > > classification at all. In the Pixel Classification applet, there > > is already a layer for Segmentation of all labels (lower left > > corner). This one just performs a simple segmentation, taking the > > most probable label. If you have a different threshold in mind, or > > want to use hysteresis thresholding for your data, you can do the > > thresholding step, and then right-click on the "Final Output" > > layer and export it. > > > > Let me know if that's what you had in mind, I'll try to walk you > > through it if it still doesn't work. > > > > Best, > > Anna > > > > > > On 12.09.2014 14:18, Joao Medeiros wrote: > > > > I will try to be more specific: I have a 3D volume and want to > > segment an object inside it (there are only 1 object). > >> > >> > >> I'm trying to use "Pixel Classification + Object Classification", > >> so I was able to: > >> > >> > >> 1) Input Data > >> 2) Feature selection > >> 3) Training > >> 4) Thresholding > >> 5) Object Feature Selection > >> > >> > >> but in Object Classification I only can add the labels (on left > >> panel) but can't select on the images (as I did in the training), > >> how to advance from this point? > >> > >> > >> In version 0.5 from training I can get the segmentation, how to > >> have something like that in 1.1 version? > >> > >> > >> Kind regards, > >> Joao Medeiros > >> > >> > >> > >> > >> > >> Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros > >> > escreveu: > >> > >> > >> > >> I'm a user from Ilastik 0.5 and now I'm trying to use the new > >> version, but the things are very different. > >> > >> > >> For example, I was able to load a stack from tiff images and I'm > >> trying to segment the image, but I can' t find how to. > >> Please someone have a walk through to segment a image from a stack? > >> > >> > >> Kind regards, > >> Joao Medeiros > >> _______________________________________________ > >> ilastik-user mailing list > >> ilastik-user at ilastik.org > >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > >> > >> > >> > >> > >> > >> _______________________________________________ > > ilastik-user mailing list ilastik-user at ilastik.org > > > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From amjoao at yahoo.com.br Mon Sep 15 18:18:20 2014 From: amjoao at yahoo.com.br (Joao Medeiros) Date: Mon, 15 Sep 2014 09:18:20 -0700 Subject: [ilastik-user] from ilastik 0.5 to 1.1 In-Reply-To: <5416F33D.8070202@iwr.uni-heidelberg.de> References: <1410520979.20415.YahooMailNeo@web164903.mail.bf1.yahoo.com> <1410524324.89804.YahooMailNeo@web164906.mail.bf1.yahoo.com> <541305F8.6020206@iwr.uni-heidelberg.de> <1410536567.98744.YahooMailNeo@web164901.mail.bf1.yahoo.com> <20140913103931.gsrxaj4u0ws08kcg@wwwmail.urz.uni-heidelberg.de> <1410769844.45625.YahooMailNeo@web164905.mail.bf1.yahoo.com> <5416F33D.8070202@iwr.uni-heidelberg.de> Message-ID: <1410797900.77083.YahooMailNeo@web164905.mail.bf1.yahoo.com> Thanks Anna, I hope the new ilastik version will be useful. The next question is: how to export the segmented object as a wavefrom OBJ mesh file (we had this option in 0.5 version, but I can't find in the new version)? Kind regards, Joao Em Segunda-feira, 15 de Setembro de 2014 15:10, Anna Kreshuk escreveu: Hi Joao, Ah, now I get it! You are absolutely right, the little 3D display is not working in the new version. The reason is that we now support "infinite" datasets and we can't make an infinite (or even very large) 3D object. We could restrict it to the field of view, like we do with the probability map, but it's not done yet. The only workflow, where this feature is enabled is Carving, which assumes that data fits in RAM. I hope ilastik can still be useful for you without this feature. Best, Anna On 15.09.2014 10:30, Joao Medeiros wrote: Hi Anna, > > >Thanks for your help and I'm sorry for so many basic questions, but it isn't working yet. I've attached a screenshot showing the question. > > >Kind regards, >Joao > > > > > > > >Em S?bado, 13 de Setembro de 2014 9:39, "anna.kreshuk at iwr.uni-heidelberg.de" escreveu: > > > >Dear Joao, > >I've made a screenshot of the Training applet, which shows where to >click to see the segmentation. Either activate all segmentations at >the same time by checking the upper checkbox, or you can do it label >by label, by clicking on the little eye icons. If it shows just one >uniform label color, it means you haven't trained the classifier yet, >so just press "live update". > >Let me know if it doesn't work for you or if you have any other >problems transitioning to the new version. > >Cheers, >Anna > > >Quoting Joao Medeiros : > >> Thank you Anna, >> >> I've trained the data. How to get a preview of the segmented object? >> In the training screen there are four panels (3 for slices and 1 >> for the 3D visualization, as in the 0.5 version) how to view the >> segmented object there? >> >> Kind regards, >> Joao >> >> >> >> Em Sexta-feira, 12 de Setembro de 2014 15:41, Anna Kreshuk >> escreveu: >> >> >> >> Hi Joao, >> >> If you only need to segment 1 object, you don't need object >> classification at all. In the Pixel Classification applet, there >> is already a layer for Segmentation of all labels (lower left >> corner). This one just performs a simple segmentation, taking the >> most probable label. If you have a different threshold in mind, or >> want to use hysteresis thresholding for your data, you can do the >> thresholding step, and then right-click on the "Final Output" >> layer and export it. >> >> Let me know if that's what you had in mind, I'll try to walk you >> through it if it still doesn't work. >> >> Best, >> Anna >> >> >> On 12.09.2014 14:18, Joao Medeiros wrote: >> >> I will try to be more specific: I have a 3D volume and want to >> segment an object inside it (there are only 1 object). >>> >>> >>> I'm trying to use "Pixel Classification + Object Classification", >>> so I was able to: >>> >>> >>> 1) Input Data >>> 2) Feature selection >>> 3) Training >>> 4) Thresholding >>> 5) Object Feature Selection >>> >>> >>> but in Object Classification I only can add the labels (on left >>> panel) but can't select on the images (as I did in the training), >>> how to advance from this point? >>> >>> >>> In version 0.5 from training I can get the segmentation, how to >>> have something like that in 1.1 version? >>> >>> >>> Kind regards, >>> Joao Medeiros >>> >>> >>> >>> >>> >>> Em Sexta-feira, 12 de Setembro de 2014 12:24, Joao Medeiros >>> escreveu: >>> >>> >>> >>> I'm a user from Ilastik 0.5 and now I'm trying to use the new >>> version, but the things are very different. >>> >>> >>> For example, I was able to load a stack from tiff images and I'm >>> trying to segment the image, but I can' t find how to. >>> Please someone have a walk through to segment a image from a stack? >>> >>> >>> Kind regards, >>> Joao Medeiros >>> _______________________________________________ >>> ilastik-user mailing list >>> ilastik-user at ilastik.org >>> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >>> >>> >>> >>> >>> >>> _______________________________________________ >> ilastik-user mailing list ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From a3perez at ucsd.edu Tue Sep 16 02:12:29 2014 From: a3perez at ucsd.edu (Alex Perez) Date: Mon, 15 Sep 2014 17:12:29 -0700 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: References: <54130552.2000702@iwr.uni-heidelberg.de> <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> Message-ID: Hi Stuart and Anna, Thank you! I was successfully able to run train_headless.py in combination with headless pixel classification to generate probability maps. One note, in my original label data I had background pixels set to a value of zero and object pixels set to a value of one. I set NUM_LABEL_CLASSES = 1, but training did not work in this way. So I changed the labels such that the background had pixels of value one and the object had pixels of value two, and this worked fine with NUM_LABEL_CLASSES = 2. Not sure if this is a bug or I just misunderstood the script. In any case, thanks a lot for the assistance! Best, Alex On Mon, Sep 15, 2014 at 6:18 AM, Berg, Stuart wrote: > Hi Alex, > > We will be releasing v1.1.2 very soon, which contains the > train_headless.py script Anna referred to. The website hasn't been updated > yet, but the Linux binary is already accessible via this link: > > http://files.ilastik.org/ilastik-1.1.2-Linux.tar.gz > > You must execute the train_headless.py script using the special python > environment shipped with ilastik. Here's how: > > $ cd ilastik-1.1.2-Linux/ > $ ./ilastik_python.sh src/ilastik/ilastik/bin/train_headless.py > MyNewProject.ilp MyRawData.png MyLabelData.png > > But before you try it, you should edit that script by hand to set the > desired input file paths, feature selections, and specify the number of > label classes in your data. (Note that label 0 is background. Class > labels are 1,2,3...) > > By the way, besides .png, you can load 3D stacks or hdf5 volumes, e.g.: > > $ ./ilastik_python.sh src/ilastik/ilastik/bin/train_headless.py > MyNewProject.ilp MyData.h5/raw MyData.h5/labels > $ ./ilastik_python.sh src/ilastik/ilastik/bin/train_headless.py > MyNewProject.ilp "mystack/raw*.png" "mystack/labels*.png" > > Best, > Stuart > > From: Alex Perez > Date: Saturday, September 13, 2014 6:46 PM > To: Anna Kreshuk > Cc: "ilastik-user at ilastik.org" > Subject: Re: [ilastik-user] Importing labels for pixel classification > > Dear Anna, > > The "train_headless.py" script is not in the bin directory of my > installation. I installed version 1.1.1 for Linux 64-bit. Thanks for your > help with the new binaries. > > Thanks, > Alex > > On Sat, Sep 13, 2014 at 1:43 AM, > wrote: > >> Dear Alex, >> >> Could you check if there is a script, named "train_headless.py" in the >> bin subdirectory of your ilastik installation? If not, you'd have to wait a >> couple days more and we'll generate all the new binaries with it. If yes, >> just try to use it and write back to us if it doesn't work or doesn't cover >> your use case. >> >> Cheers, >> Anna >> >> >> Quoting Alex Perez : >> >> Dear Anna, >>> >>> Yes, I am interested in trying non-interactive training in that way. That >>> would be great. Thanks! >>> >>> Best, >>> Alex >>> On Sep 12, 2014 7:38 AM, "Anna Kreshuk" >> heidelberg.de> >>> wrote: >>> >>> Dear Alex, >>>> >>>> It is not possible to load your own labels interactively, but you can >>>> train like that in headless mode. Would you be interested in such >>>> non-interactive training? I could send you a script for it then. >>>> >>>> Best, >>>> Anna >>>> >>>> On 04.09.2014 20:30, Alex Perez wrote: >>>> >>>> Dear All, >>>> >>>> I have a stack of binary label images that I would like to import as >>>> labels into Ilastik's Pixel Classification workflow. Is this possible? >>>> >>>> I am using Ilastik version 1.1.1. >>>> >>>> Thank you, >>>> Alex >>>> >>>> >>>> >>>> >>>> _______________________________________________ >>>> ilastik-user mailing listilastik-user at ilastik.orghttp:// >>>> mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >>>> >>>> >>>> >>>> >>> >> >> >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: From bergs at janelia.hhmi.org Tue Sep 16 02:34:50 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Tue, 16 Sep 2014 00:34:50 +0000 Subject: [ilastik-user] Importing labels for pixel classification In-Reply-To: References: <54130552.2000702@iwr.uni-heidelberg.de> <20140913104343.80adtkux44kgkoc0@wwwmail.urz.uni-heidelberg.de> Message-ID: An HTML attachment was scrubbed... URL: From bolek.zapiec at gen.mpg.de Thu Sep 25 12:52:20 2014 From: bolek.zapiec at gen.mpg.de (Bolek Zapiec) Date: Thu, 25 Sep 2014 10:52:20 +0000 Subject: [ilastik-user] Ilastik classifier for CellProfiler Message-ID: <0D56EBF9-E98F-46CA-86AC-BBD017AB42D0@gen.mpg.de> Hi everyone, I?ve been using the ilastik for a while now but just recently took notice of the big 1.0 release and since 0.5 was so great I figured its certainly worth a shot. My workflow in the past has been to use ilastik produced classifiers using the ClassifyPixels module in CellProfiler where I later refined the classifications with some morph functions and object filters. These are performed on images on the order of 15,000x15,000 pixels. (the data set ultimately constitutes an XYZ dataset but since they?re serial fluorescent cryosections the segmentation transpires only in XY for now) Alas after making a new project with the new version I am at a loss as to how I can export the classifier for use in CellProfiler. I?ve also tried batch processing a set of these images within ilastik but only managed to get an output of either probabilities or segmentation in h5 files which I can?t seem to get cellProfiler to load for further processing. I?ve gone through all the documentation online and read back through this mailing list but can?t seem to figure out what I?m missing. Do I have to go back to 0.5 to get the option to export my classifier for use in CP? Is there a performance benefit to use Ilastik first then continue processing in CP? Thanks! Bolek From Benjamin.Pavie at cme.vib-kuleuven.be Fri Sep 26 13:50:16 2014 From: Benjamin.Pavie at cme.vib-kuleuven.be (Benjamin Pavie) Date: Fri, 26 Sep 2014 11:50:16 +0000 Subject: [ilastik-user] Exporting carving segmentation result Message-ID: Dear, I am using Ilastik 1.1.2 using carving and was wandering how I could export the result of the carving segmentation. The only option I see is to export the mesh obj, but how can I export the mask so I can edit it manually like we can do with Pixel Classification for example? Benjamin -------------- next part -------------- An HTML attachment was scrubbed... URL: From bergs at janelia.hhmi.org Fri Sep 26 15:46:19 2014 From: bergs at janelia.hhmi.org (Berg, Stuart) Date: Fri, 26 Sep 2014 13:46:19 +0000 Subject: [ilastik-user] Exporting carving segmentation result Message-ID: An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: 58F161F6-B426-446F-AD60-7A4F379B7D12.png Type: image/png Size: 75429 bytes Desc: 58F161F6-B426-446F-AD60-7A4F379B7D12.png URL: From Benjamin.Pavie at cme.vib-kuleuven.be Fri Sep 26 16:30:16 2014 From: Benjamin.Pavie at cme.vib-kuleuven.be (Benjamin Pavie) Date: Fri, 26 Sep 2014 14:30:16 +0000 Subject: [ilastik-user] Exporting carving segmentation result In-Reply-To: References: Message-ID: Thanks for your quick answer, I didn't see this option. I also realized that I had to change the order of export from txyzc to tzxyc and then it worked perfectly! Thanks Benjamin From: Berg, Stuart [mailto:bergs at janelia.hhmi.org] Sent: vendredi 26 septembre 2014 15:46 To: Benjamin Pavie; ilastik-user at ilastik.org Subject: Re: [ilastik-user] Exporting carving segmentation result Hi Benjamin, Any of the layers listed on the left-hand side of the screen can be exported by right-clicking the layer item in the list. - To export the object you are currently editing, export the "Segmentation" layer. - To export all saved objects as a binary image, export the "Completed segments (unicolor)" layer. All pixels in the exported image will have a value of 0 or 1. - To export all saved objects as a label image, export the "Completed segments (one color per object)" layer. Pixels will have values of 0 (background) or 1,2,3,... for each segmented object. Overlapping objects will be clipped. Best regards, Stuart [cid:image001.png at 01CFD9A7.316BFC00] From: Benjamin Pavie > Date: Friday, September 26, 2014 7:50 AM To: "ilastik-user at ilastik.org" > Subject: [ilastik-user] Exporting carving segmentation result Dear, I am using Ilastik 1.1.2 using carving and was wandering how I could export the result of the carving segmentation. The only option I see is to export the mesh obj, but how can I export the mask so I can edit it manually like we can do with Pixel Classification for example? Benjamin -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image001.png Type: image/png Size: 75429 bytes Desc: image001.png URL: From Benjamin.Pavie at cme.vib-kuleuven.be Fri Sep 26 16:37:02 2014 From: Benjamin.Pavie at cme.vib-kuleuven.be (Benjamin Pavie) Date: Fri, 26 Sep 2014 14:37:02 +0000 Subject: [ilastik-user] 3D Measurement Message-ID: Dear All, I used Ilastik to segment my EM images and now I would like to do some measurement. Which tools here do you recommend? I would especially want to measure the surface contact between 2 3D objects. Is there an open source solution that can do it? I tried 3D slicer which is very useful to do the 3D rendering but couldn't figure a way to do surface contact measurement between 2 objects. Thanks, Benjamin -------------- next part -------------- An HTML attachment was scrubbed... URL: From fred.hamprecht at iwr.uni-heidelberg.de Fri Sep 26 22:13:57 2014 From: fred.hamprecht at iwr.uni-heidelberg.de (Fred Hamprecht) Date: Fri, 26 Sep 2014 22:13:57 +0200 Subject: [ilastik-user] 3D Measurement In-Reply-To: References: Message-ID: <5425C905.40207@iwr.uni-heidelberg.de> Hello Benjamin, I have heard that Graham Knott (cc'ed) is developing an analysis tool based on Blender. He may be willing to share a pre-release version. Best regards, Fred On 26-Sep-14 16:37, Benjamin Pavie wrote: > Dear All, > > I used Ilastik to segment my EM images and now I would like to do some > measurement. Which tools here do you recommend? > > I would especially want to measure the surface contact between 2 3D > objects. > > Is there an open source solution that can do it? > > I tried 3D slicer which is very useful to do the 3D rendering but > couldn?t figure a way to do surface contact measurement between 2 objects. > > Thanks, > > Benjamin > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From anna.kreshuk at iwr.uni-heidelberg.de Sat Sep 27 08:31:03 2014 From: anna.kreshuk at iwr.uni-heidelberg.de (anna.kreshuk at iwr.uni-heidelberg.de) Date: Sat, 27 Sep 2014 08:31:03 +0200 Subject: [ilastik-user] Ilastik classifier for CellProfiler In-Reply-To: <0D56EBF9-E98F-46CA-86AC-BBD017AB42D0@gen.mpg.de> References: <0D56EBF9-E98F-46CA-86AC-BBD017AB42D0@gen.mpg.de> Message-ID: <20140927083103.44osb36isgwsgsss@wwwmail.urz.uni-heidelberg.de> Hi Bolek, We started to port the ilastik to CellProfiler bridge, but I believe this work is not finished yet. Cheers, Anna Quoting Bolek Zapiec : > Hi everyone, > > I?ve been using the ilastik for a while now but just recently took > notice of the big 1.0 release and since 0.5 was so great I figured > its certainly worth a shot. > > My workflow in the past has been to use ilastik produced classifiers > using the ClassifyPixels module in CellProfiler where I later > refined the classifications with some morph functions and object > filters. These are performed on images on the order of 15,000x15,000 > pixels. (the data set ultimately constitutes an XYZ dataset but > since they?re serial fluorescent cryosections the segmentation > transpires only in XY for now) > > Alas after making a new project with the new version I am at a loss > as to how I can export the classifier for use in CellProfiler. I?ve > also tried batch processing a set of these images within ilastik but > only managed to get an output of either probabilities or > segmentation in h5 files which I can?t seem to get cellProfiler to > load for further processing. I?ve gone through all the documentation > online and read back through this mailing list but can?t seem to > figure out what I?m missing. Do I have to go back to 0.5 to get the > option to export my classifier for use in CP? Is there a performance > benefit to use Ilastik first then continue processing in CP? > > Thanks! > Bolek > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > > From christoph.sommer at imba.oeaw.ac.at Mon Sep 29 09:03:51 2014 From: christoph.sommer at imba.oeaw.ac.at (Christoph Sommer) Date: Mon, 29 Sep 2014 09:03:51 +0200 Subject: [ilastik-user] Ilastik classifier for CellProfiler In-Reply-To: <20140927083103.44osb36isgwsgsss@wwwmail.urz.uni-heidelberg.de> References: <0D56EBF9-E98F-46CA-86AC-BBD017AB42D0@gen.mpg.de> <20140927083103.44osb36isgwsgsss@wwwmail.urz.uni-heidelberg.de> Message-ID: <54290457.3020908@imba.oeaw.ac.at> Hi Bolek, I already wrote a CellProfiler plug-in for ilastik 1.x. The name of the module is "ilastik_pixel_classification" and it is part of the current CellProfiler release. This module is very similar to the 0.5 one, but uses an ilastik pixel classification project file as input. However, ilastik 1.x is not yet included in the CellProfiler binaries / installers. Are you working on a CellProfiler source installation? Best, Christoph > > Hi Bolek, > > We started to port the ilastik to CellProfiler bridge, but I believe > this work is not finished yet. > > Cheers, > Anna > > > Quoting Bolek Zapiec : > >> Hi everyone, >> >> I?ve been using the ilastik for a while now but just recently took >> notice of the big 1.0 release and since 0.5 was so great I figured >> its certainly worth a shot. >> >> My workflow in the past has been to use ilastik produced classifiers >> using the ClassifyPixels module in CellProfiler where I later >> refined the classifications with some morph functions and object >> filters. These are performed on images on the order of 15,000x15,000 >> pixels. (the data set ultimately constitutes an XYZ dataset but >> since they?re serial fluorescent cryosections the segmentation >> transpires only in XY for now) >> >> Alas after making a new project with the new version I am at a loss >> as to how I can export the classifier for use in CellProfiler. I?ve >> also tried batch processing a set of these images within ilastik but >> only managed to get an output of either probabilities or >> segmentation in h5 files which I can?t seem to get cellProfiler to >> load for further processing. I?ve gone through all the documentation >> online and read back through this mailing list but can?t seem to >> figure out what I?m missing. Do I have to go back to 0.5 to get the >> option to export my classifier for use in CP? Is there a performance >> benefit to use Ilastik first then continue processing in CP? >> >> Thanks! >> Bolek >> >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >> > > > > _______________________________________________ > ilastik-user mailing list > ilastik-user at ilastik.org > http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user > From bolek.zapiec at gen.mpg.de Mon Sep 29 11:10:07 2014 From: bolek.zapiec at gen.mpg.de (Bolek Zapiec) Date: Mon, 29 Sep 2014 09:10:07 +0000 Subject: [ilastik-user] Ilastik classifier for CellProfiler In-Reply-To: <54290457.3020908@imba.oeaw.ac.at> References: <0D56EBF9-E98F-46CA-86AC-BBD017AB42D0@gen.mpg.de> <20140927083103.44osb36isgwsgsss@wwwmail.urz.uni-heidelberg.de> <54290457.3020908@imba.oeaw.ac.at> Message-ID: <069691BE-019A-40D9-8234-98F12658D1F2@gen.mpg.de> Hi Christoph, I?ve got the CellProfiler 2.12 Trunk running on Windows 8.1 (support states windows 8 but 8.1 is necessary for 512GB RAM) - I just had a look and saw the new ?ilastik_Pixel_classification? module but it says ?IlastikPixelClassification cannot run on this platform because the necessary libraries are not avaialble. IlastikPixelClassification is supported on 64-bit versions ov vista, 7 and windows 8, and on Linux. I do have ilastik installed on the machine (1.1.1) so I presume I must manually install the ilastik binaries into CP somehow. Having once compiled a CellProfiler with ilastik for a linux cluster I can imagine this may be more than a 3 step process, but if there are instructions somewhere on how I could proceed with this I?d be pretty grateful. Thanks, Bolek On Sep 29, 2014, at 9:03 AM, Christoph Sommer wrote: > Hi Bolek, > > I already wrote a CellProfiler plug-in for ilastik 1.x. The name of the module is "ilastik_pixel_classification" and it is part of the current CellProfiler release. This module is very similar to the 0.5 one, but uses an ilastik pixel classification project file as input. > > However, ilastik 1.x is not yet included in the CellProfiler binaries / installers. Are you working on a CellProfiler source installation? > > Best, > Christoph > >> >> Hi Bolek, >> >> We started to port the ilastik to CellProfiler bridge, but I believe this work is not finished yet. >> >> Cheers, >> Anna >> >> >> Quoting Bolek Zapiec : >> >>> Hi everyone, >>> >>> I?ve been using the ilastik for a while now but just recently took notice of the big 1.0 release and since 0.5 was so great I figured its certainly worth a shot. >>> >>> My workflow in the past has been to use ilastik produced classifiers using the ClassifyPixels module in CellProfiler where I later refined the classifications with some morph functions and object filters. These are performed on images on the order of 15,000x15,000 pixels. (the data set ultimately constitutes an XYZ dataset but since they?re serial fluorescent cryosections the segmentation transpires only in XY for now) >>> >>> Alas after making a new project with the new version I am at a loss as to how I can export the classifier for use in CellProfiler. I?ve also tried batch processing a set of these images within ilastik but only managed to get an output of either probabilities or segmentation in h5 files which I can?t seem to get cellProfiler to load for further processing. I?ve gone through all the documentation online and read back through this mailing list but can?t seem to figure out what I?m missing. Do I have to go back to 0.5 to get the option to export my classifier for use in CP? Is there a performance benefit to use Ilastik first then continue processing in CP? >>> >>> Thanks! >>> Bolek >>> >>> _______________________________________________ >>> ilastik-user mailing list >>> ilastik-user at ilastik.org >>> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >>> >>> >> >> >> >> _______________________________________________ >> ilastik-user mailing list >> ilastik-user at ilastik.org >> http://mailman.ilastik.org/cgi-bin/mailman/listinfo/ilastik-user >> >