v1.0.0 - AAAI 2025 Official Release

PoseMamba v1.0.0

AAAI 2025 official release — documentation, issue templates, and pretrained weight links.

Highlights

• PoseMamba-L: 38.1 mm MPJPE on Human3.6M (P1, detected 2D), 6.7M params, 27.9G MACs
• Bidirectional spatio-temporal Mamba/SSM for monocular 3D human pose estimation
• In-the-wild demo via vis.py

What's in this release

• Rewritten README (SOTA table, Quick Start, FAQ)
• Installation & FAQ Discussion (pinned)
• Issue templates (Bug Report / Question)
• Fixed config paths: configs/pose3d/
• Correct evaluation: --evaluate best_epoch.bin (see eval.sh)

Pretrained weights

Model	Params	Download
PoseMamba-S	0.9M	Google Drive
PoseMamba-B	3.4M	Google Drive
PoseMamba-L	6.7M	Google Drive
All-in-one	—	Bundle

Hugging Face mirrors coming soon. Track progress in repo README.

Quick start

conda create -n posemamba python=3.8.5 && conda activate posemamba
pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 --extra-index-url https://download.pytorch.org/whl/cu117
pip install -r requirements.txt
cd kernels/selective_scan && pip install -e . && cd ../..
python vis.py --video sample_video.mp4 --gpu 0

Paper

• AAAI 2025
• arXiv:2408.03540

Citation

@inproceedings{huang2025posemamba,
  title={PoseMamba: Monocular 3D Human Pose Estimation with Bidirectional Global-Local Spatio-Temporal State Space Model},
  author={Huang, Yunlong and Liu, Junshuo and Xian, Ke and Qiu, Robert Caiming},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2025}
}