Sports2D: Compute 2D human pose and angles from a video or a webcam

Quick release for a dirty fix to the OpenMMLab servers having shut down on 06/06/2026. Mirrored the models I had stored locally to https://huggingface.co/datasets/DavidPagnon/rtmlib_models

[!IMPORTANT] Request: If anyone can find any of the missing models, please tell me or submit a pull request! On Windows, they should be located at C:\Users\<YOUR_USERNAME>\.cache\rtmlib\hub\checkpoints (replace <YOUR_USERNAME>)

List of changes:

• OpenMMLab servers are down: Uploaded the ones I have locally stored to Hugging-face and monkey-patched the corrected URLs. Some are still missing (list below). Will submit a pull request to RTMlib asap.
• Fixed wrong filtering key in post-processing
• Improved website: added filming and storage guidelines, responsive on mobile, call to visit website on Readme, etc
• Created an issue template to help me categorize problems
• Switched from toml to rtoml, from os.path to pathlib, imageio_ffmpeg to av

Still missing:

- rtmo-l_16xb16-600e_body7-640x640-b37118ce_20231211
- rtmpose-l_simcc-body7_pt-body7-halpe26_700e-256x192-2abb7558_20230605
- rtmpose-l_simcc-body7_pt-body7_420e-256x192-4dba18fc_20230504
- rtmpose-l_simcc-body7_pt-body7_420e-384x288-3f5a1437_20230504
- rtmpose-l_simcc-ucoco_dw-ucoco_270e-256x192-4d6dfc62_20230728
- rtmpose-l_simcc-ucoco_dw-ucoco_270e-384x288-2438fd99_20230728
- rtmpose-m_simcc-face6_pt-in1k_120e-256x256-72a37400_20230529
- rtmpose-m_simcc-ucoco_dw-ucoco_270e-256x192-c8b76419_20230728
- rtmpose-s_simcc-face6_pt-in1k_120e-256x256-d779fdef_20230529
- rtmpose-s_simcc-ucoco_dw-ucoco_270e-256x192-3fd922c8_20230728
- rtmpose-t_simcc-body7_pt-body7-halpe26_700e-256x192-6020f8a6_20230605
- rtmpose-t_simcc-face6_pt-in1k_120e-256x256-df79d9a5_20230529
- rtmpose-t_simcc-ucoco_dw-ucoco_270e-256x192-dcf277bf_20230728
- yolox_l_8xb8-300e_humanart-ce1d7a62
- yolox_m_8xb8-300e_humanart-c2c7a14a
- rtmw-dw-m-s_simcc-cocktail14_270e-256x192_20231122
- rtmw-x_simcc-cocktail13_pt-ucoco_270e-384x288-0949e3a9_20230925

Successfully reuploaded:

- rtmo-m_16xb16-600e_body7-640x640-39e78cc4_20231211
- rtmo-s_8xb32-600e_body7-640x640-dac2bf74_20231211
- rtmdet_nano_8xb32-300e_hand-267f9c8f
- rtmpose-m_simcc-ap10k_pt-aic-coco_210e-256x256-7a041aa1_20230206
- rtmpose-m_simcc-body7_pt-body7-halpe26_700e-256x192-4d3e73dd_20230605
- rtmpose-m_simcc-body7_pt-body7_420e-256x192-e48f03d0_20230504
- rtmpose-m_simcc-hand5_pt-aic-coco_210e-256x256-74fb594_20230320
- rtmpose-s_simcc-body7_pt-body7-halpe26_700e-256x192-7f134165_20230605
- rtmpose-s_simcc-body7_pt-body7_420e-256x192-acd4a1ef_20230504
- rtmpose-t_simcc-body7_pt-body7_420e-256x192-026a1439_20230504
- rtmpose-x_simcc-body7_pt-body7-halpe26_700e-384x288-7fb6e239_20230606
- rtmpose-x_simcc-body7_pt-body7_700e-384x288-71d7b7e9_20230629.zip  (1
- yolox_m_8xb8-300e_humanart-c2c7a14a
- yolox_nano_8xb8-300e_humanart-40f6f0d0
- yolox_s_8xb8-300e_humanart-3ef259a7
- yolox_tiny_8xb8-300e_humanart-6f3252f9
- yolox_x_8xb8-300e_humanart-a39d44ed
- rtmw-dw-x-l_simcc-cocktail14_270e-256x192_20231122
- rtmw-dw-x-l_simcc-cocktail14_270e-384x288_20231122

Also uploaded

For safety reasons, I also uploaded all the other models listed on the rtmlib repo, even if they were not hosted on the openmmlab's server.

- RTMW3D/rtmw3d-x_8xb64_cocktail14-384x288-b0a0eab7_20240626.onnx
- YOLOX/yolox_nano.onnx
- YOLOX/yolox_tiny.onnx
- YOLOX/yolox_s.onnx
- YOLOX/yolox_m.onnx
- YOLOX/yolox_l.onnx
- YOLOX/yolox_darknet.onnx
- YOLOX/yolox_x.onnx
- easy_ViTPose/vitpose-s-coco.onnx
- easy_ViTPose/vitpose-b-coco.onnx
- easy_ViTPose/vitpose-l-coco.onnx
- easy_ViTPose/vitpose-s-coco_25.onnx
- easy_ViTPose/vitpose-b-coco_25.onnx
- easy_ViTPose/vitpose-l-coco_25.onnx
- easy_ViTPose/vitpose-s-wholebody.onnx
- easy_ViTPose/vitpose-b-wholebody.onnx
- easy_ViTPose/vitpose-l-wholebody.onnx
- easy_ViTPose/vitpose-s-apt36k.onnx
- easy_ViTPose/vitpose-b-apt36k.onnx
- easy_ViTPose/vitpose-h-apt36k.onnx

Sources:

• https://github.com/Tau-J/rtmlib
• https://openmmlab.com
• https://huggingface.co/Soykaf/RTMW3D-x
• https://github.com/Megvii-BaseDetection/YOLOX/releases/tag/0.1.1rc0
• https://huggingface.co/JunkyByte/easy_ViTPose

Dataset Detail

I only uploaded models that were listed with an Apache 2.0 license (or compatible, such as (MIT, BSD-3 Clause...)

Full Changelog: https://github.com/davidpagnon/Sports2D/compare/v0.8.31...v0.8.32