Armand Collin

For some reason imageio 2.26 is downloaded, which makes all test fail. Despite the pinned `imageio

> Likely a loose version condition in one of the pip dependencies that we install in the conda env file - those won't enfoce our own conditions (annoyingly, and we...

As of right now the segment CLI works (only with image inputs ATM, not folders).

A comment about this new `prepare_input()` function: https://github.com/axondeepseg/axondeepseg/blob/98885ff8a1b73d93e0869edd82799e8b4ed08f25/AxonDeepSeg/segment.py#L103-L123 This code is bad... We have 2 blocks: lines 105-114 deal with incorrect nb. of channels and lines 116-123 deal with wrong...

Regarding my last comment, I think 1b42dfa fixes everything. The function is now much more readable and the processing is more gracious.

As of 7293d0e, the segmentation CLI is now fully functional :tada: :tada: :tada: :tada: --- A comment about API changes in segment.py - `segment_image()` is now `segment_images()` (with plural) -...

@mathieuboudreau are you using conda? I used mamba for my env and yes I remember it took some time. Regarding the tests, yes that's a good idea. I wanted to...

During inference tests, I'm getting weird runtime errors but only on ubuntu (works fine on macos and windows). ``` Error: mkl-service + Intel(R) MKL: MKL_THREADING_LAYER=INTEL is incompatible with libgomp-a34b3233.so.1 library....

About my last comment: in the end, I had to hardcode an environment variable for the `ubuntu-latest` gh actions. But the tests now pass on all platforms :champagne: We should...

> Hi @hermancollin - what's the state of this PR, and where should I focus my efforts to help you get this to the finish line? @mathieuboudreau the main things...