The principle and algorithm of deeplabcut?

Hello, I have achieved an unmarked pose estimate of a horse using deeplabcut, which has a very friendly experience.But I don’t want to just use it, I want to know deeplabcut’s algorithm and principle of every step, and why you choose this algorithm.For example, how does deeplabcut use DeeperCut’s feature detectors----a deep ResNet?
I have read your paper, but there is no relevant explanation in your paper.Do you have any other articles about deeplabcut’s principle?
I hope you enjoy my questions and look forward to your reply.

Hey Liuxin, Thanks for your interest. I’m not sure what your question is. You can read the code here and see how it is used: https://github.com/AlexEMG/DeepLabCut/blob/master/deeplabcut/pose_estimation_tensorflow/predict_videos.py

Cheers,
Alexander