Help for extracting color image features

[songxinyuty]

hello,I want to extract geometric topological features of color images(RGB three channels) by giotto-tda, but the example for images of handwritten digits have one channel. Can you give me a suggestion about how I can use giotto-tda to extract color image features?

[ulupo]

Thanks for getting in touch! I think this depends a lot on your application and I don't think a "best" method is known or obvious. I can help you with the API once you narrow down a way you'd like to approach this, but don't have the time right now to do the research myself, as I don't have much experience with color image data.

[ektas0330]

Perhaps it would be useful to continue this in the discussion section, but commenting here as this is still open.
@songxinyuty, I had a similar question when I was working on using TDA for color images. Though I did not directly generate persistence diagrams from color images, I found a way to use giotto-tda for my application (where inputs were color images). Here's what I did in case you want to take a look. But as @ulupo mentioned, it would depend on what problem you are trying to solve by using TDA for color images.

[ulupo]

Thanks @ektas0330! Sorry for getting back late. I've moved this to a discussion as it is not an issue with the package but still of interest. I hope some people land here and find the link to your paper helpful! You would also be welcome to contribute this recipe as a jupyter notebook example, if you wish.