KL-Loss icon indicating copy to clipboard operation
KL-Loss copied to clipboard

Any plans for PyTorch version with MMDetection?

Open LiJunnan1992 opened this issue 4 years ago • 7 comments

Hi Yihui,

This method is hugely interesting and I would like to incorporate it in my research. Do you have any plans to release a PyTorch version of the code? It would be very useful to the community.

Thank you!

LiJunnan1992 avatar Sep 13 '19 08:09 LiJunnan1992

@LiJunnan1992 Do you incorporated successfully?

LittleBoy7 avatar Sep 23 '19 07:09 LittleBoy7

@LittleBoy7 Not yet...

LiJunnan1992 avatar Sep 23 '19 10:09 LiJunnan1992

@LittleBoy7 @LiJunnan1992 Hi, I reimplemented kl-loss in with my own project Stronger-yolo-pytorch. You may find some help here.

apxlwl avatar Nov 10 '19 15:11 apxlwl

@wlguan This is awesome! I'll add a link to your repo in the readme

ethanhe42 avatar Nov 25 '19 15:11 ethanhe42

@wlguan I want to reimplemented KL loss , but I cannot find your project( Stronger-yolo-pytorch), could you please give me some help?

lucs-C avatar Oct 09 '20 08:10 lucs-C

@wlguan I want to reimplemented KL loss , but I cannot find your project( Stronger-yolo-pytorch), could you please give me some help?

here's a fork https://github.com/yihui-he/Stronger-yolo-pytorch

ethanhe42 avatar Oct 23 '20 04:10 ethanhe42

Hi, I have implemented a pytorch version of KL-Loss with MMDetection in my project KL-Loss-pytorch. The reproduced results are as follows:

KL Loss Var Vote soft-NMS AP (Original Paper) AP (Reproduced)
:x: :x: :x: 37.9 38.4
:heavy_check_mark: :x: :x: 38.5 39.2
:heavy_check_mark: :heavy_check_mark: :x: 38.8 39.8
:heavy_check_mark: :heavy_check_mark: :heavy_check_mark: 39.2 40.2

cxliu0 avatar Aug 29 '22 03:08 cxliu0