add multi-stage loss #204
add multi-stage loss #204
Conversation
|
|
mmpose/models/detectors/top_down.py
Outdated
| output[-1][target_weight.squeeze(-1) > 0].unsqueeze( | ||
| 0).detach().cpu().numpy(), | ||
| target[target_weight.squeeze(-1) > 0].unsqueeze( | ||
| 0).detach().cpu().numpy()) | ||
| else: | ||
| _, avg_acc, cnt = pose_pck_accuracy( |
There was a problem hiding this comment.
Nope, I think. These cnts have been here in previous versions. Should I change the cnts into _ ?
| 0).detach().cpu().numpy()) | ||
| if target.dim() == 5 and target_weight.dim() == 4: | ||
| _, avg_acc, cnt = pose_pck_accuracy( | ||
| output[-1][target_weight[:, -1, :, :].squeeze(-1) > 0]. |
There was a problem hiding this comment.
question: why is the last output special
There was a problem hiding this comment.
Generally speaking, the last output is the final prediction (used for calculating accuracy), while the intermediate outputs are used for training.
There was a problem hiding this comment.
cool, @wusize may add a line of comment explaining this magic number
|
the main change is ok, but there are some unintended changes |
Codecov Report
@@ Coverage Diff @@
## master #204 +/- ##
==========================================
+ Coverage 82.75% 82.90% +0.14%
==========================================
Files 106 106
Lines 6189 6211 +22
Branches 994 1001 +7
==========================================
+ Hits 5122 5149 +27
+ Misses 877 874 -3
+ Partials 190 188 -2
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
|
Hi @wusize First of all, we want to express our gratitude for your significant PR in the MMpose project. Your contribution is highly appreciated, and we are grateful for your efforts in helping improve this open-source project during your personal time. We believe that many developers will benefit from your PR. We would also like to invite you to join our Special Interest Group (SIG) private channel on Discord, where you can share your experiences, ideas, and build connections with like-minded peers. To join the SIG channel, simply message moderator— OpenMMLab on Discord or briefly share your open-source contributions in the #introductions channel and we will assist you. Look forward to seeing you there! Join us ：https://discord.gg/UjgXkPWNqA If you have WeChat account，welcome to join our community on WeChat. You can add our assistant ：openmmlabwx. Please add "mmsig + Github ID" as a remark when adding friends：） |
* add multi-stage loss * Modifications
* add multi-stage loss * Modifications
add multi-stage loss