[Feature] Support Human-Art Dataset

[juxuan27]

merge dev-1.x to main

Motivation

This PR is to add Human-Art dataset into MMPose.

Modification

The modifications are listed as follows:

• Add configs of Human-Art in the folder "configs" (The download links still need to be updated, at present, only td-hm_hrnet-w48_8xb32-210e_humanart-256x192 is trained. However, since Human-Art is a multi-scenario dataset that has the same keypoint definition as MSCOCO, users can use checkpoint trained on MSCOCO to test generalization ability of different models on Human-Art. And that is why I think giving different configs of Human-Art is necessary although the models are still not ready.)
• Edit documents of README.md, README_CN.md, docs/en/dataset_zoo/2d_body_keypoint.md, and docs/zh_cn/dataset_zoo/2d_body_keypoint.md (I find docs/zh_cn/dataset_zoo/2d_body_keypoint.md is in fact write in English hhh. So I follow previous doc add in English)
• Add Human-Art in mmpose/datasets/datasets/body/init.py, mmpose/datasets/datasets/body/humanart_dataset.py, configs/base/datasets/humanart.py, and configs/base/datasets/humanart_aic.py. Noted that the official usage of Human-Art is joint training with MSCOCO, so I did not write the training config in dataset combine mode (define two datasets and use one data config to combine them together), but simply use the json file we provided in our google drive, training_humanart_coco.json.

BC-breaking (Optional)

It does not.

Use cases (Optional)

It is the same with MSCOCO. E.G. Users can simply convert configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_hrnet-w48_8xb32-210e_coco-256x192.py to configs/body_2d_keypoint/topdown_heatmap/humanart/td-hm_hrnet-w48_8xb32-210e_humanart-256x192.py for training and testing on Human-Art.

Checklist

Before PR:

• I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• 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:

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

[CLAassistant]

All committers have signed the CLA.

[Tau-J]

Thank you for your contribution.
Would you mind simplifying the submitted config files? For example, you can only keep the important ckpts used in your paper experiments or choose a few representative algorithms to support Human-Art.
Also, I noticed that the sample images you provided in tests/data/humanart are from COCO dataset. Please replace them with the unique images from human-art. Thank you very much!

[juxuan27]

Thank you for your comments. I will delete the unnecessary config files and change the corresponding images and annotation for testing. I will also check for the reason of failed unit tests.

[codecov]

Codecov Report

Patch coverage: 56.73% and project coverage change: -2.82 ⚠️

Comparison is base (43397c3) 82.06% compared to head (b23421f) 79.24%.

❗ Current head b23421f differs from pull request most recent head 5b143d9. Consider uploading reports for the commit 5b143d9 to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #2304      +/-   ##
===========================================
- Coverage    82.06%   79.24%   -2.82%     
===========================================
  Files          232      249      +17     
  Lines        13643    14865    +1222     
  Branches      2319     2578     +259     
===========================================
+ Hits         11196    11780     +584     
- Misses        1914     2507     +593     
- Partials       533      578      +45     

Flag	Coverage Δ
unittests	79.24% <56.73%> (-2.82%)	⬇️

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

Impacted Files	Coverage Δ
mmpose/codecs/regression_label.py	87.09% <ø> (ø)
mmpose/models/heads/__init__.py	100.00% <ø> (ø)
mmpose/models/heads/coord_cls_heads/rtmcc_head.py	93.61% <ø> (ø)
mmpose/models/heads/coord_cls_heads/simcc_head.py	90.26% <ø> (ø)
mmpose/models/pose_estimators/base.py	81.08% <0.00%> (-2.26%)	⬇️
mmpose/models/utils/rtmcc_block.py	72.86% <ø> (+0.21%)	⬆️
mmpose/structures/bbox/transforms.py	68.70% <ø> (ø)
mmpose/structures/keypoint/__init__.py	100.00% <ø> (ø)
mmpose/structures/keypoint/transforms.py	93.93% <ø> (-0.18%)	⬇️
mmpose/visualization/__init__.py	100.00% <ø> (ø)
... and 34 more

... and 3 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[Tau-J]

LGTM