How to build key points for foot recognition

I want to build a foot keypoint recognition, similar to the existing hand recognition, with 21 keypoints.
But I don’t know how to start building it, I have downloaded the code for hand recognition and I would like to do it by changing the layout of the keypoints, but most of them are imported from an off-the-shelf library and I can’t get started.
Any suggestions please?
Software: Pycharm ; System: macOs

that alone wont work, you need to train your own network here.
you can probably reuse the openpose training code or their dataset

also, just curious, what do you want to do ?
and you probably meant: foot detection, not recognition ?

pose estimation (foot has this orientation in space), distinct from classification (“that’s (not) a picture of a foot”, “that’s a left/right foot”), localization (“where’s Waldo”), recognition (“I’ve seen that identical foot before”). “detection” is a fuzzy word, most often means localization, sometimes means classification (binary, there/not-there). felt like splitting hairs just now.