you’ll find some references in the original pr
Zheng F, Deng C, Sun X, et al. Pyramidal person re-identification via multi-loss dynamic training[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 8514-8522.
Yu F, Jiang X, Gong Y, et al. Devil’s in the Details: Aligning Visual Clues for Conditional Embedding in Person Re-Identification[J]. arXiv e-prints, 2020: arXiv: 2009.05250.
in the end, there a pretrained cnn, you process images through it , and receive a simple 1d feature vector, that can be easily compared with L2 or cosine norm .