Mailman 3 November 2015 - scikit-image

Postdoc training at the University of Washington in neuroengineering/data science
by Ariel Rokem 10 Nov '15

10 Nov '15

With apologies for cross-posting, I am posting the following on behalf of my colleague, Ione Fine: Two excellent postdoctoral fellowship opportunities with a deadline of January 15th are available at the University of Washington, Seattle, USA: http://uwin.washington.edu/post-docs/apply-post-docs/ http://escience.washington.edu/postdoctoral-fellowships Candidates with a strong computational background (e.g., machine learning, computer vision, neuroengineering, etc.) are sought to work on the following project: Prof. Fine has over several years worked in collaboration with Second Sight <http://www.secondsight.com/> (developers of a retinal prosthetic, analogous to a cochlear implant, on the market). She has developed developed a model that, for any given pulse train, is pretty good at predicting what a patient implanted with a retinal prosthetic will see (essentially a linear-nonlinear model with some weird tweaks because the retina is responding to current instead of light). But what the field really needs is the *reverse* of this model – we need to be able to predict what electrical pulses (across the set of electrodes) will produce a percept that most closely matches the percept that would normally be elicited by whatever it is the patient is looking at. It’s actually a really tricky problem for a variety of reasons. Building such a model would be of very high impact on the field, because it wouldn’t just help Second Sight patients – it would likely be generalized by all the other groups trying to build prosthetic devices (e.g., with optogenetics). Please contact Prof. Fine (ionefine(a)uw.edu) if you are interested in this particular project or just want information about UWIN ( http://uwin.washington.edu/). Feel free to also contact me (arokem(a)gmail.com) for questions about the Data Science Environment at the University of Washington (http://escience.washington.edu/) and the eScience fellowships.

1 0

Re: Edge completion/linking algorithm in Python?
by Juan Nunez-Iglesias 06 Nov '15

06 Nov '15

Update: it's a mix of C++ and Matlab, licensed as GPL. Good times. =P On Fri, Nov 6, 2015 at 3:32 PM Juan Nunez-Iglesias <jni.soma(a)gmail.com> wrote: > Hey Robin! > > Coincidentally, I was just yesterday going through this paper, "Projection > onto the Manifold of Elongated Structures for Accurate Extraction": > http://infoscience.epfl.ch/record/211536/files/top.pdf > > Their source code is available here: > > https://documents.epfl.ch/groups/c/cv/cvlab-unit/www/src/NN_projections_12.… > > I actually don't even know what language they used because I'm travelling > and on a crappy connection, so the 65MB package is just chugging along at > 33KBps. =P I presume it's C++, based on previous software from that group, > but it might be easily wrappable in Cython. > > I'm not aware of other options at the moment, sorry! > > Juan. > > On Fri, Nov 6, 2015 at 7:28 AM Robin Wilson <r.t.wilson.bak(a)googlemail.com> > wrote: > >> Hi, >> >> Does anyone on this list know of an implementation of some sort of edge >> linking or edge completion algorithm in Python? I've got some edges, >> produced with the skimage implementation of canny, with gaps in them which >> I'm finding hard to fill. I've implemented various 'hacky' methods myself, >> but I wondered if anyone knew of any code already available to do this in >> Python? >> >> Cheers, >> >> Robin >> >> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/d/optout. >> >

1 0

Edge completion/linking algorithm in Python?
by Robin Wilson 06 Nov '15

06 Nov '15

Hi, Does anyone on this list know of an implementation of some sort of edge linking or edge completion algorithm in Python? I've got some edges, produced with the skimage implementation of canny, with gaps in them which I'm finding hard to fill. I've implemented various 'hacky' methods myself, but I wondered if anyone knew of any code already available to do this in Python? Cheers, Robin

2 1

Re: [ANN] colorspacious 1.0.0 released
by Stefan van der Walt 03 Nov '15

03 Nov '15

Hi Nathaniel On 2015-11-03 01:38:31, Nathaniel Smith <njs(a)vorpus.org> wrote: > I just released version 1.0.0 of colorspacious, a library for > converting between colorspaces in Python, and thought that > scikit-image might be interested. This is the library that we used to > design viridis and several other new colormaps, and it includes common > well-known colorspaces (sRGB, LAB, XYZ, xyY) along with several more > sophisticated models for estimating perceptual correlates (CIECAM02), > estimating perceptual similarity (CAM02-UCS), and simulating > colorblindness, all wrapped up in a very easy-to-use interface. > (There's basically just one function.) Do you think it would make sense to turn colorspacious into a dependency and rely on it for our existing conversions? Are there any speed implications? Stéfan

1 0

hough ellipse detection
by Matthias Verstraete 03 Nov '15

03 Nov '15

Hy, I'm trying to detect ellipses using the hough_ellipse method provided by scikit-image. However, one of the axis of the result ellipse is always zero. I've tries adjusting all the parameters but nothing helped. Does anyone know why only one axis can be larger than 0 and how I can fix it? Thanks

2 1

[ANN] colorspacious 1.0.0 released
by Nathaniel Smith 03 Nov '15

03 Nov '15

Hi all, I just released version 1.0.0 of colorspacious, a library for converting between colorspaces in Python, and thought that scikit-image might be interested. This is the library that we used to design viridis and several other new colormaps, and it includes common well-known colorspaces (sRGB, LAB, XYZ, xyY) along with several more sophisticated models for estimating perceptual correlates (CIECAM02), estimating perceptual similarity (CAM02-UCS), and simulating colorblindness, all wrapped up in a very easy-to-use interface. (There's basically just one function.) Notable features of this release include a fancy new tutorial and reference manual: https://colorspacious.readthedocs.org/ and 100% test coverage. Downloads: https://pypi.python.org/pypi/colorspacious/ Source code: https://github.com/njsmith/colorspacious Share and enjoy! -n -- Nathaniel J. Smith -- http://vorpus.org

1 0

Strange behaviour - Normalised Cut (on Macbook)
by bricklemacho＠gmail.com 03 Nov '15

03 Nov '15

Hi All, Running this example: http://scikit-image.org/docs/dev/auto_examples/plot_ncut.html#example-plot-… On one machine, example works as expected. On my work machine I am getting the following results of normalised cuts example: http://imgur.com/uWmvW2p Details on machines and what I have tried is below. Any help appreciated. Regards, Michael. -- Macbook 1 (work as expected): -------------------------- OS X Yosmite 10.10.4 Python 2.7.9 install via mac ports skimage.__version__ 0.11.3 install via mac ports Macbook 2 (strange behaviour) -------------------------- OS X Yosmite 10.10.5 Python 2.7.10 installed via mac ports skimage.__verison__ 0.11.3 installed via mac ports Things I have tried: 1. Reinstall from source via macports sudo rm -rf /opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/skimage* sudo rm -rf /opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/scikit_image* sudo port upgrade -s -n --force py27-scikit-image skimage.__version__: 0.11.3, identical result (http://imgur.com/uWmvW2p) 2. Install latest development verison sudo rm -rf /opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/skimage* sudo rm -rf /opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/scikit_image* git clone https://github.com/scikit-image/scikit-image.git cd skikit-image python setup.py build sudo python setup.py install skimage.__version__: 0.12dev, identical result: http://imgur.com/uWmvW2p

2 3

Feature freeze and release
by Stefan van der Walt 02 Nov '15

02 Nov '15

Hi, everyone I would like to propose a feature freeze so that we can release a new version of scikit-image on 11/11/2015. Please review your favorite issues and pull requests in preparation: https://github.com/scikit-image/scikit-image/issues https://github.com/scikit-image/scikit-image/pulls The release milestone is here: https://github.com/scikit-image/scikit-image/pulls?q=is%3Aopen+is%3Apr+mile… Stéfan

1 0