Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

detection 复现 #38

Open
luckin99 opened this issue May 16, 2023 · 5 comments
Open

detection 复现 #38

luckin99 opened this issue May 16, 2023 · 5 comments

Comments

@luckin99
Copy link

作者您好!我尝试在mmdetection的框架下复现您在fasterRCNN的实验,但结果一直不同,能求您分享一下FasterRCNNr101-r50的log吗?
同时我有一些小的细节问题。我看论文里说是对proposal的cls结果进行kd,但代码中是对rpn head 再过roi_head里的bbox_head得到的结果。第二是这里有用inherit吗?
谢谢!!!

@Zzzzz1
Copy link
Collaborator

Zzzzz1 commented May 17, 2023

rpn head的输出就是proposal,再经过bbox_head就是该proposal的cls结果呀

@luckin99
Copy link
Author

那我的理解应该没错,就是bbox输出的大小为(1024,81)的logits,但我在复现您这里fasterRCNN-FPN的R101-R50的实验的时候,KD和DKD都出现了性能下降。KD为-1 mAP,dkd为-0.5 mAP,多次尝试都无法复现。您这有log吗,我想对比一下loss尺度可以吗

@luckin99
Copy link
Author

或者这里面有什么训练的trcik吗

@Zzzzz1
Copy link
Collaborator

Zzzzz1 commented May 17, 2023

log确实已经找不到了,时间比较久远。。
复现是指用我们开源的这个代码训det,但是点数对不上吗?
我们当时用的detectron2版本的commit-id是:221448e4dcfbebf215b8d21ae7e4b1dfbf422d29
不知道对齐版本后能否解决?

@Zzzzz1
Copy link
Collaborator

Zzzzz1 commented May 17, 2023

或者这里面有什么训练的trcik吗

没什么额外的训练trick,用这份代码应该是能跑出来paper中汇报的结果的。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants