isl-org
Turbo colormap
Hi, first thanks for this nice paper!
For plotting in this repo, what would you think of using a different colormap designed for depth? It make it much easier to see details, compare distant points in the image, and is friendlier to color blindness. Also it has already been added to matplotlib a while back so is widely available. Subjectively I think it also highlight the quality of your work! (it's only a detail I know, but a bit of a pet peeve of mine 😅)
More sample images
is friendlier to color blindness
Turbo is not colorblind-friendly, nor is any other rainbow colormap; see this blog post for details. The existing Magma colormap is much better for accessibility.
@mpetroff thanks for the write-up, good to know! What do you think of the other two points? Do you happen to know a colormap that successfully address all three concerns?
Unfortunately, I think colorblind accessibility and maximizing perceptual distance for typical color vision, and thus addressing your other two points, are mutually exclusive goals.
To maintain accessibility, you need to rely primarily on lightness differences in the colormap, which limits the achievable perceptual distance between difference parts of the colormap. By also relying on hue, as is done by Turbo, you can increase the average perceptual distance but only for individuals that perceive hue the same way. Referencing the plots in my blog post, Turbo is better for your first two points for individuals with typical color vision because the ΔE metric is higher than, e.g., Viridis, but as the metric is highly variable for individuals with color-vision deficiencies, Turbo is not colorblind accessible.