jiyanggao / Video-Person-ReID

Video-based Person ReID Method Implementations on MARS
381 stars 110 forks source link

About the ResNet50TA network #1

Open kikyou123 opened 6 years ago

kikyou123 commented 6 years ago

Hi, I run the model ResNet50TA in Mars dataset. And I get the performance CMC: 83.2, 93.5, 95.7, 97.1; map: 76.3. But I find the attention seems not work, because the learned score for every frame is 0.25.

lzj322 commented 6 years ago

@kikyou123 There is a typo here https://github.com/jiyanggao/Video-Person-ReID/blob/dfe9a02e9dcae1a7aef66d9fd6a7896f015c6394/models/__init__.py#L7 It's actually pooling code. I think it should be corrected.@jiyanggao

LongWeiZJU commented 6 years ago

I also find this problem. I have corrected the obvious bug pointed by @lzj322. But still in most cases, the attention score for every frame is 0.25. @jiyanggao