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
Attempt to Reproduce the Results of CondInst. #39
Comments
@Yuxin-CV Unfortunately, we use design B ... |
Thanks for your reply. |
BTW, will the code of CondInst be released recently? I found it is really hard for me to reproduce the results... |
@Yuxin-CV Our paper is in submission, so we won't release the code until our paper gets accepted. You may refer to other implementation of CondInst. |
Thanks for your reply. |
I am also interested in the loss behavior of FCOS part in CondInst, e.g., cls_loss & reg_loss. Compared with FCOS, is the loss become higher or lower in CondInst? The information is quite helpful for me to debug. Thanks! |
@Yuxin-CV I cannot find the log files now, but I think the detector's losses should be lower because of the improved detection performance. |
Thanks for your suggestion~ |
Hi~ @tianzhi0549 The
The results in different Resolution of Mask Prediction shows similar & reasonable Box AP (39.5), but the Mask AP is abnormal, especially for the 1/4 resolution case. So I think at least there are some problems in the alignment of mask feature during training (I use the to
The masks then rescale to the original image resolution (using Also, there is a 0.2 AP gap for Box(#20 (comment)). This indicates that there must be some problems in the training code, probably in the feature alignment of mask prediction & GT (but I already use the So I wonder could you @tianzhi0549 please help me with the above problems? |
@Yuxin-CV We have released the code of BlendMask. CondInst is implemented with the same codebase. I think it should be helpful to you. Also, a hint is if the performance degradation is due to misalignment, you should see much more performance degradation on small objects than on large objects. |
Thanks for your suggestions. @tianzhi0549
It seems that now the bottleneck is not APs (misalignment) for the 1 / 2 case.
Could you please give me some suggestions? |
BTW, I want to make sure that:
Thanks! @tianzhi0549 |
@Yuxin-CV 1) The mask branch should be similar to the basis module in BlendMask. But we do not upsample the feature maps from 8x to 4x here. I don't think these design choices of the mask branch are critical. 2) Yes. |
Thanks for your reply! |
Hi~ @tianzhi0549, thanks for your reply. |
|
Hi~ @tianzhi0549
I want to make sure the shared head architecture of CondInst.
Design A
Design B
Which one is right?
I found Design B will degradation Box AP and mask AP is also very low.
Here is my results for MS-R-50_1x.
Box AP
Mask AP
The Box AP should be higher than 39.5 for MS training(~39.5) & multi-task training(+~1.0). So I think Design B is wrong. It is hard for one branch to handle 3 preds, and the grad from controller_pred degenerate the reg_pred.
The text was updated successfully, but these errors were encountered: