About Generative module and GAN loss

chenhao2345 / GCL

Implementation for CVPR2021 paper "Joint Generative and Contrastive Learning for Unsupervised Person Re-identification"

MIT License

46 stars 12 forks source link

About Generative module and GAN loss #10

Open YangJae96 opened 3 years ago

YangJae96 commented 3 years ago

The above is the generative module in the paper.

Can I ask why you applied the x'org part in the GAN loss?

Is it because the training generator is hard without using x'org part?

chenhao2345 commented 2 years ago

Yes, it helps to train generator. This part is traditional reconstruction in GAN and AE. Reconstruction allows for better adapting the decoder to the person distribution.