switch band order to scarlet band order

[ismael-mendoza]

This is more of a selfish change, I've always wanted the order of the images shape to be [n_bands, nx, ny] rather than [nx, ny, n_bands]. This is also the convention used in scarlet and in bliss so it would be nice to have everything consistent and not have to remember to transpose (at least for me). This also simplifies the code in plot_utils.py slightly.

Would this cause any problems for anyone? Did I miss any places?

[codecov]

Codecov Report

Merging #113 (0000a1a) into main (20d1631) will increase coverage by 0.05%.
The diff coverage is 92.85%.

@@            Coverage Diff             @@
##             main     #113      +/-   ##
==========================================
+ Coverage   74.57%   74.63%   +0.05%     
==========================================
  Files          10       10              
  Lines         948      954       +6     
==========================================
+ Hits          707      712       +5     
- Misses        241      242       +1     

Flag	Coverage Δ
unittests	74.63% <92.85%> (+0.05%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
btk/draw_blends.py	90.90% <88.88%> (-0.26%)	⬇️
btk/measure.py	84.74% <100.00%> (ø)
btk/plot_utils.py	46.15% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 20d1631...0000a1a. Read the comment docs.

[thuiop]

Well I do usually have the channels as the last dimension... maybe we should have a tag defining the order and doing the transposition automatically ?

[ismael-mendoza]

Well I do usually have the channels as the last dimension... maybe we should have a tag defining the order and doing the transposition automatically ?

Yeah I can do that, I think I will use this order as the default for everything, but add an optional keyword argument to the draw_blend_generator so that it can return the images transposed in a specific way. Would that work for you?

[ismael-mendoza]

OK I just added a keyword dim_order, let me know what you think @thuiop

[thuiop]

I think it works that way. Tensorflow uses a string with values "NHWC" or "NCHW" as a tag (for channel last and channel first), this may be more user-friendly ? I don't think anyone would want to have the channels as the middle one anyway. But it's a very minor thing anyway.

[thuiop]

See my last comment ; apart from that it seems OK for me, as long as the tests run it should be working.

[ismael-mendoza]

I think it works that way. Tensorflow uses a string with values "NHWC" or "NCHW" as a tag (for channel last and channel first), this may be more user-friendly ? I don't think anyone would want to have the channels as the middle one anyway. But it's a very minor thing anyway.

Ah I like this idea better actually, I will quickly make the change and then merge when tests are passing. Thanks!