Emmanuelle Gouillart

Hi @alemirone why would you want to force the positivity of the output?

Here are the results from benchmarking all functions for which check_parallel returns a finite depth (warning; there are still some bugs, there are still functions in the list which should...

And here is the same table on the BIDS machine which has a lot of cores (32). Thank you @yuvipanda! function | execution time | apply_parallel time | ratio --...

The script for running this benchmark is https://github.com/emmanuelle/skimage-sprint/blob/master/benchmark_dask_simplified.py (it uses a pickled list of suitable functions that should be downloaded as well). Unfortunately the code only runs with **this branch**.

As requested on gitted by @stsievert, here is the benchmarking result using joblib (on my laptop, 4 logical cores). And the script is https://github.com/emmanuelle/skimage-sprint/blob/master/benchmark_joblib.py function | execution time | apply_parallel_joblib...

@jni random walker was written to be insensitive to image magnitude, the ``beta`` parameter is normalized by ``img.std()`` but I just saw that this is a bug, ``beta`` should be...

Thanks a lot for the PR @liudj2008 ! The CI failures probably don't have anything to do with your PR: on py2.7, it looks like a new version of qtconsole...

One difference between RGB and HSL is that the bounds for the channels are different: - for RGB the 3 channels are expected to be between 0 and 255 -...

For the `build-doc` CI issue, could you please update your master branch, and then merge master into your feature branch to see if this solves the problem?

@nicolaskruchten we could indeed take the HSL input and convert it back to rgb to make a png, but then we're not making use of the HSL capability of the...