๐ŸŽจ color mapping

lo(image_01).rgb

plt.imshow(image_11[:,:,1], cmap="twilight", vmin=-1); # One single channel.

This works, but now this would be interpreted as a 1960 196x3 RGBA images.

What I had in ming was more like 30 196x196 RGBA images (3 channels for each of the 10 images).

Keep this in mind when using cmap.

image_batch = image_01.transpose(2,0,1)[None].repeat(10, axis=0)
print(lo(image_batch))

vals = (image_batch + 1)/2
lut_idxs = (vals * cmap.N).astype(np.int64)

mapped = lut.take(lut_idxs, axis=0, mode="clip")

print(lo(mapped))
lo(mapped[:2]).rgb # First 2 of the images, each as 3 channels.
array[10, 3, 196, 196] n=1152480 xโˆˆ[-4.053e-09, 1.000] ฮผ=0.361 ฯƒ=0.248
array[10, 3, 196, 196, 4] n=4609920 xโˆˆ[0.067, 1.000] ฮผ=0.534 ฯƒ=0.334

Extend the lut to cover +/-inf too.

source

InfCmap

 InfCmap (cmap:matplotlib.colors.Colormap, below:Optional[str]=None,
          above:Optional[str]=None, nan:Optional[str]=None,
          ninf:Optional[str]=None, pinf:Optional[str]=None)

*Matplotlib colormap extended to have colors for +/-inf

Parameters extept cmap are matplotlib color strings.*

Type Default Details
cmap Colormap Base matplotlib colormap
below Optional None Values below 0
above Optional None Values above 1
nan Optional None NaNs
ninf Optional None -inf
pinf Optional None +inf 
tcmap = InfCmap(get_cmap("twilight"),
                  below="blue", above="red", nan="yellow")
rgb(tcmap(bad_image[:,:,0])) # Note: Mapped only first channel

tcmap = InfCmap(get_cmap("twilight"),
                  below="blue", above="red",
                  nan="yellow", ninf="cyan", pinf="fuchsia")
rgb(tcmap(bad_image)[:,:,0]) # Note: Mapped all channels, show only the mapping for the first.

tcmap = InfCmap(get_cmap("bwr"),
                  below="blue", above="red",
                  nan="yellow", ninf="cyan", pinf="fuchsia")
rgb(tcmap(bad_image.transpose(2,0,1)))