Mailman 3 October 2011 - scikit-image

Re: A bit of history and the bilateral filter
by Stéfan van der Walt Oct. 14, 2011

Oct. 14, 2011

On Thu, Oct 13, 2011 at 8:06 PM, Tony Yu <tsyu80(a)gmail.com> wrote: > Just for reference, there's also an implementation of the bilateral filter > in cellprofiler: > > https://svn.broadinstitute.org/CellProfiler/trunk/CellProfiler/cellprofiler… > > I haven't really looked into either of these, so I'm in no way suggesting > that one is better than the other. The implementation I referred to is just a naïve version, without any optimisation; I know it can be done more efficiently. We should benchmark the two, at least. Cheers Stéfan

1 0

Re: A bit of history and the bilateral filter
by Tony Yu Oct. 13, 2011

Oct. 13, 2011

2011/10/13 Stéfan van der Walt <stefan(a)sun.ac.za> > Hi all > > I just came across a paper in which the author uses the bilateral > filter. This reminded me that we have an implementation lying around > somewhere; and it so happens that we discussed it in the very e-mail > that led to the formation of this scikit :) > > http://comments.gmane.org/gmane.comp.python.numeric.general/28316 > > Strangely, though, the bilateral filter never made its way in! If you > are new to scikits-image and would like to make a contribution, this > is an excellent opportunity: > > https://github.com/stefanv/bilateral/tree/master/bilateral > Just for reference, there's also an implementation of the bilateral filter in cellprofiler: https://svn.broadinstitute.org/CellProfiler/trunk/CellProfiler/cellprofiler… I haven't really looked into either of these, so I'm in no way suggesting that one is better than the other. Best, -Tony

1 0

A bit of history and the bilateral filter
by Stéfan van der Walt Oct. 13, 2011

Oct. 13, 2011

Hi all I just came across a paper in which the author uses the bilateral filter. This reminded me that we have an implementation lying around somewhere; and it so happens that we discussed it in the very e-mail that led to the formation of this scikit :) http://comments.gmane.org/gmane.comp.python.numeric.general/28316 Strangely, though, the bilateral filter never made its way in! If you are new to scikits-image and would like to make a contribution, this is an excellent opportunity: https://github.com/stefanv/bilateral/tree/master/bilateral Regards Stéfan

1 0

Challenge: estimate a blurring kernel
by Stéfan van der Walt Oct. 13, 2011

Oct. 13, 2011

Hi all, At a recent Adobe MAX conference, people went crazy over their deblurring demo [1]. I want it in scikits-image! The authors were even kind enough to leave us some clues as to how [2]. Who is up for the challenge? Stéfan [1] http://www.youtube.com/watch?v=xxjiQoTp864t [2] http://people.csail.mit.edu/sparis/publi/2011/cvpr_radon/Cho_11_Blur_Kernel…

1 0

Re: Skeletonization
by Tony Yu Oct. 13, 2011

Oct. 13, 2011

2011/10/13 Stéfan van der Walt <stefan(a)sun.ac.za> > Hi Neil, > > On Thu, Oct 13, 2011 at 6:45 AM, Neil Yager <yager.neil(a)gmail.com> wrote: > > I'm thinking about doing a thresholding algorithm next, followed by > > perhaps gabor filters or gray-level co-occurance matrices (if no one > > is already working on these). > > I've got some old grey-level c.o.m. code here, if you want to have a look: > > http://mentat.za.net/cgi-bin/hgwebdir.cgi/greycomatrix/ > > Adaptive / auto-thresholding should be fun! > > Also, cellprofiler has some thresholding code available: https://svn.broadinstitute.org/CellProfiler/trunk/CellProfiler/cellprofiler… The "contributions" page on the skimage website only mentions converting the morphology and filter modules from cellprofiler, but I thinking everything in the cpmath module is BSD licensed. I played around with the thresholding code a bit: it was a bit slow and the interface was a little confusing, but it may serve as a good starting point. Best, -Tony

1 0

Re: Skeletonization
by Stéfan van der Walt Oct. 13, 2011

Oct. 13, 2011

Hi Neil, On Thu, Oct 13, 2011 at 6:45 AM, Neil Yager <yager.neil(a)gmail.com> wrote: > I really like the looks of this project and I'm hoping to add a few > things. However, I've never contributed to open source before, so I'll > definitely need some guidance along the way. Welcome! We look forward to your contributions. > I've written some code to perform thinning on binary objects (i.e. > skeletonization). I've created a file that contains the logic, another > file with an example of usage, and one for unit tests. It would be very interesting to compare what you've done with the skeletonization code already in progress. We tried to merge it just before the release, but since it was part of another, massive patch, it didn't make it. Have a look here: http://bit.ly/ski_skeleton > 1. Which package is best for this function? I've put it in the > 'morphology' package, but I'm not sure it really belongs there because > I don't use any morphological operations. (note: this implementation > should be significantly faster than a morphology-based approach) We had the other patch in morphology too, so that's probably as good a place as any. > 2. Unit test: I've written a few tests checking IO, but I can't see > how to write tests to check if it is creating the "correct" skeleton, > because there isn't really a "correct" skeleton. There's lots of > thinning algorithms out there, and they often give different results > (e.g. the medial axis transform gives different skeletons than this > implementation). How does one write useful unit tests in these cases? In this case, it's quite reasonable to construct a test case, thin it with your algorithm, and inspect the result to ensure that it is correct. You then store that result (e.g. in a .npz or .png file), and use it in the comparison. > 3. github: I've never used git before, and I'm still getting used to > the terminology. What is the best way for someone to review my code? > I've now pushed it to a branch (neil_yager-skeletonize) in my repo > (NeilYager). If you were to test it out, would you fork it or pull it? You previously made a successful pull request against the main repository--that's exactly the right way to go. We review the pull requests on GitHub, and leave comments inline. > I'm thinking about doing a thresholding algorithm next, followed by > perhaps gabor filters or gray-level co-occurance matrices (if no one > is already working on these). I've got some old grey-level c.o.m. code here, if you want to have a look: http://mentat.za.net/cgi-bin/hgwebdir.cgi/greycomatrix/ Adaptive / auto-thresholding should be fun! Regards Stéfan

1 0

Skeletonization
by Neil Yager Oct. 13, 2011

Oct. 13, 2011

Hi everyone, I really like the looks of this project and I'm hoping to add a few things. However, I've never contributed to open source before, so I'll definitely need some guidance along the way. I've written some code to perform thinning on binary objects (i.e. skeletonization). I've created a file that contains the logic, another file with an example of usage, and one for unit tests. A few questions: 1. Which package is best for this function? I've put it in the 'morphology' package, but I'm not sure it really belongs there because I don't use any morphological operations. (note: this implementation should be significantly faster than a morphology-based approach) 2. Unit test: I've written a few tests checking IO, but I can't see how to write tests to check if it is creating the "correct" skeleton, because there isn't really a "correct" skeleton. There's lots of thinning algorithms out there, and they often give different results (e.g. the medial axis transform gives different skeletons than this implementation). How does one write useful unit tests in these cases? 3. github: I've never used git before, and I'm still getting used to the terminology. What is the best way for someone to review my code? I've now pushed it to a branch (neil_yager-skeletonize) in my repo (NeilYager). If you were to test it out, would you fork it or pull it? In general, any comments about coding style/efficiency/documentation/ etc would be very welcome. I'm thinking about doing a thresholding algorithm next, followed by perhaps gabor filters or gray-level co-occurance matrices (if no one is already working on these). Cheers, Neil

1 0

Re: A big thanks for 0.3!
by Stéfan van der Walt Oct. 13, 2011

Oct. 13, 2011

Hi Nelle On Wed, Oct 12, 2011 at 10:30 PM, Nelle Varoquaux <nelle.varoquaux(a)gmail.com> wrote: > I think that the "merge" button disappears when github can't do the merge > without manual intervention, so you have to do it "by hand", by fetching the > code, and rebasing locally on your computer. > Else, I don't think the pull request themselves are going to be affected. Of > course, some of the patches will need some reworking. I was concerned that the directory renaming from 'scikits/image' to 'skimage' would make trouble, but maybe Git can track that. > I worked on that yesterday, and I almost have a patch running. > Unfortunately, I seem to have a problem with the cython generated part, that > I can't import (I can't find the *.so files manually as well). I have the > same problem on master. I'm using cython 0.15.1. > As soon as I've sorted that cython problem, I'll have a patch. Can you elaborate on this problem? Maybe we can help. Stéfan

1 0

Re: A big thanks for 0.3!
by Nelle Varoquaux Oct. 13, 2011

Oct. 13, 2011

2011/10/10 Stéfan van der Walt <stefan(a)sun.ac.za> > Hi Nelle > > On Mon, Oct 10, 2011 at 11:51 AM, Nelle Varoquaux > <nelle.varoquaux(a)gmail.com> wrote: > > I can do that, if it can wait until this week-end! I've used grep in the > > past for such tasks :) > > Thanks, that'd be great! Could you also investigate what happens to > our pull requests when we change the naming? I have a feeling they > may not appreciate it. > I think that the "merge" button disappears when github can't do the merge without manual intervention, so you have to do it "by hand", by fetching the code, and rebasing locally on your computer. Else, I don't think the pull request themselves are going to be affected. Of course, some of the patches will need some reworking. > The scikit-learn guys have both scikits.learn and sklearn subpackages, > but we'll stick to one and simply move from scikits.image to skimage > for the next release. > I worked on that yesterday, and I almost have a patch running. Unfortunately, I seem to have a problem with the cython generated part, that I can't import (I can't find the *.so files manually as well). I have the same problem on master. I'm using cython 0.15.1. As soon as I've sorted that cython problem, I'll have a patch. > > Other low hanging fruit would be to improve the documentation index: > > http://scikits-image.org/docs/ > > (It's currently not themed along with the rest of the website) > > Then there's also the compatibility table, where I want to list all > the functionality we provide: > > http://scikits-image.org/docs/dev/coverage_table.html > > At the moment it mainly contains MATLAB features. I think that was a > mistake on my part; we should not follow MATLAB, but rather forge our > own exciting route ahead (playing catch-up is boring). > > Lots more to be done, but I'll stop here :) > > Regards > Stéfan >

1 0

Re: Bug-fix 0.3.1
by Stéfan van der Walt Oct. 12, 2011

Oct. 12, 2011

Hi, On Wed, Oct 12, 2011 at 2:53 PM, Klonuo Umom <klonuo(a)gmail.com> wrote: > If of any interest here is output of my > test: http://pastebin.com/raw.php?i=DJ8JGtGU > I have opencv 2.3.1 in /usr/local/lib/ with cv2.so for Python in > dist-packages (on Ubuntu) The tests look good. To get opencv working, you'll have to symlink cv2.so to libcv.so. Stefan

1 0