search

NVlabs / DG-Net

:couple: Joint Discriminative and Generative Learning for Person Re-identification. CVPR'19 (Oral) :couple:
https://www.zdzheng.xyz/publication/Joint-di2019
Other
1.28k stars 228 forks source link

online feeding and identity supervision #78

Open Wooks-git opened 1 year ago

Wooks-git commented 1 year ago

Hi, author. I have a question about your paper. You used online feeding and identity supervision to get the high quality of the generated image. Is it Primary feature learning and Fine-grained feature mining loss, respectively?

layumi commented 1 year ago

Yes. These two losses are for the generated images x^i_j in the paper .

Wooks-git commented 1 year ago

Thanks for your reply. Then if don't use two losses, can't generate x^i_j images? or even if generate x^i_j images, generated low quality?

layumi commented 1 year ago

Hi @Wooks-git I do not try it, since we focus on the human representation.

Yes. You can try it. I think the image quality will still be in a reasonable range.