What color model to choose for machine learning?

Question

Working with color

Hello!
There is a task - to recognize the color of clothes in the photo in a textual description of the color. There are color models: RGB, HSV, HSL and more.
But still, all these models eventually give more than 800 shades and such names as Gainsboro, blue-steel and other Avocado colors. What color model (or space) would provide an "easy way" to bind a paletted color to a textual color name? And we want to make the colors not 800, but for example 40

Answer 1

The task is not as simple as it seems. Lots of pitfalls. If, for example, you have a multi-brand store and pictures come from different sources, there will be problems with different shooting conditions (light temperature, overall brightness). Plus, many brands process their photos in Photoshop to shit, excuse my French.
And it's not clear - do you have a list of colors ready or do you need to get it from an array of photos?
Best would be CIE Lab or LCh(ab). And related CIE delta E functions.

Answer 2

Cluster colors into 40 clusters using k-means method?

Answer 3

if you already have 40 reference colors, then just look for the smallest distance to the reference colors.
HSV/HSL or LAB will give a more accurate result than rgb.