[Feature] rtmpose on AnimalPose

[wukurua]

Motivation

I train the the rtmpose model on AnimalPose dataset and add related config files and triain results. (But how can I post the ckpt and log files just like in other md files?)

Modification

configs/animal_2d_keypoint/rtmpose/animalpose/*

Checklist

Before PR:

• [x] I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• [x] Pre-commit or linting tools indicated in CONTRIBUTING.md are used to fix the potential lint issues.
• [ ] Bug fixes are covered by unit tests, the case that causes the bug should be added in the unit tests.
• [ ] New functionalities are covered by complete unit tests. If not, please add more unit tests to ensure correctness.
• [ ] The documentation has been modified accordingly, including docstring or example tutorials.

After PR:

• [x] CLA has been signed and all committers have signed the CLA in this PR.

[CLAassistant]

All committers have signed the CLA.

[Tau-J]

Thanks for your contribution! You can upload ckpts and logs(the json file) to an online driver like google, onedriver, etc. I'll save it to our server and return your a link.

[Tau-J]

BTW, it seems that your PR branch is checkout from the main branch, which lags behind the dev-1.x. Would you mind rebasing it from dev-1.x following our docs?

[codecov[bot]]

Codecov Report

Attention: 64 lines in your changes are missing coverage. Please review.

Comparison is base (4fa943b) 82.26% compared to head (c0b3dfd) 81.99%. Report is 112 commits behind head on dev-1.x.

:exclamation: Current head c0b3dfd differs from pull request most recent head 41cb971. Consider uploading reports for the commit 41cb971 to get more accurate results

Files	Patch %	Lines
mmpose/visualization/opencv_backend_visualizer.py	41.66%	48 Missing and 8 partials :warning:
mmpose/visualization/local_visualizer.py	76.47%	3 Missing and 1 partial :warning:
mmpose/apis/inferencers/base_mmpose_inferencer.py	66.66%	2 Missing :warning:
mmpose/apis/inferencers/pose2d_inferencer.py	60.00%	1 Missing and 1 partial :warning:

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #2329      +/-   ##
===========================================
- Coverage    82.26%   81.99%   -0.28%     
===========================================
  Files          232      232              
  Lines        13582    13643      +61     
  Branches      2307     2319      +12     
===========================================
+ Hits         11173    11186      +13     
- Misses        1881     1921      +40     
- Partials       528      536       +8     

Flag	Coverage Δ
unittests	81.99% <51.87%> (-0.28%)	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[Tau-J]

There is one thing that I'm concerned. The models you trained only used the RTMPose model architecture, but did not use the training strategy and data augmentation of RTMPose. These techniques are also a part of RTMPose and can bring better model performance. You can refer to the AP10K config we provided. Thank you again for your contribution but could you please modify the config and retrain the models?

[wukurua]

There is one thing that I'm concerned. The models you trained only used the RTMPose model architecture, but did not use the training strategy and data augmentation of RTMPose. These technologies are also a part of RTMPose and can bring better model performance. You can refer to the AP10K config we provided. Thank you again for your contribution but could you please modify the config and retrain the models?

Actually, I initially used the training strategy and data augmentation of RTMPose.But the final effect is not as good as it is now.

[wukurua]

Thanks for your contribution! You can upload ckpts and logs(the json file) to an online driver like google, onedriver, etc. I'll save it to our server and return your a link.

https://drive.google.com/drive/folders/1TSNWTx5tjx6fF4DnM-tgsUw2sfFmcw3f Uploaded here👆

[Tau-J]

oops, thanks for your feedback. I'll double check again on rtmpose-m. It seems that the rtmpose-m you trained is suboptimal, the performance of which is even worse than rtmpose-s.

[wukurua]

Yes, I also think this result is weird. In fact, I have trained with batch_size=32 earlier, the result is like this. Maybe I can retrain this to check ths result again.

[OpenMMLab-Assistant-004]

Hi @wukurua,

We'd like to express our appreciation for your valuable contributions to the mmpose. Your efforts have significantly aided in enhancing the project's quality. It is our pleasure to invite you to join our community thorugh Discord_Special Interest Group (SIG) channel. This is a great place to share your experiences, discuss ideas, and connect with other like-minded people. To become a part of the SIG channel, send a message to the moderator, OpenMMLab, briefly introduce yourself and mention your open-source contributions in the #introductions channel. Our team will gladly facilitate your entry. We eagerly await your presence. Please follow this link to join us: ​https://discord.gg/UjgXkPWNqA.

If you're on WeChat, we'd also love for you to join our community there. Just add our assistant using the WeChat ID: openmmlabwx. When sending the friend request, remember to include the remark "mmsig + Github ID".

Thanks again for your awesome contribution, and we're excited to have you as part of our community!