---

PSGait: Gait Recognition using Parsing Skeleton

 

Hangrui Xu, Zhengxian Wu, Chuanrui Zhang, Zhuohong Chen, Zhifang Liu, Peng Jiao, Haoqian Wang*

🔥 Highlights

• 🧠 Novel Representation: Introduces Parsing Skeleton to capture fine-grained human dynamics
• 🎯 Higher Information Entropy than traditional silhouettes or skeletons
• ⚡ Plug-and-Play: Compatible with existing gait recognition frameworks (e.g., OpenGait)
• 🚀 Performance Gain: Up to +15.7% Rank-1 accuracy improvement
• 🪶 Lightweight & Efficient: Better performance with reduced computational cost

---

📰 News

• 🔥 (2026.02) Paper is available on arXiv: PSGait: Gait Recognition using Parsing Skeleton
• 🎉 (2026.01) This paper has been accepted to ICASSP 2026.

---

👀 Abstract

Gait recognition has emerged as a robust biometric modality due to its non-intrusive nature. Conventional gait recognition methods mainly rely on silhouettes or skeletons. While effective in controlled laboratory settings, their limited information entropy restricts generalization to real-world scenarios. To overcome this, we propose a novel representation called Parsing Skeleton, which uses a skeleton-guided human parsing method to capture fine-grained body dynamics with much higher information entropy. To effectively explore the capability of the Parsing Skeleton, we also introduce PSGait, a framework that fuses Parsing Skeleton with silhouettes to enhance individual differentiation. Comprehensive benchmarks demonstrate that PSGait outperforms state-of-the-art multimodal methods while significantly reducing computational resources. As a plug-and-play method, it achieves an improvement of up to 15.7% in the accuracy of Rank-1 in various models. These results validate the Parsing Skeleton as a lightweight, effective, and highly generalizable representation for gait recognition in the wild.

---

⚙️ Environment Setup

Install Environment via Anaconda (Recommended)

conda create -n psgait python=3.9.0
conda activate psgait
pip install -r requirements.txt

---

🧩 Skeleton Generation (Optional)

For dataset samples without provided skeletons, we use SAPIENS to generate the corresponding skeletons from the images.

---

📦 Data Preparation Pipeline

Step 1: Generate Parsing Skeleton Images

You need to modify the skeleton path, silhouette path, output path, and log path to your desired locations.

parsing.sh

Step 2: Generate Aligned Images

Modify the skeleton path, silhouette path, output path, and log path to your desired locations.

transform.sh

Step 3: Convert Images to .pkl Files

The silhouette path here should point to the aligned images.

pkl_process.sh

🏃 Training

You can use the aligned pkl files to train various gait recognition networks.

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 opengait/main.py --cfgs ./configs/baseline/baseline.yaml --phase train

📜 License

This repository is released under the Apache 2.0 license.

😉 Citation

Please consider citing our paper if our code are useful:

@article{xu2025psgait,
      title={PSGait: Gait Recognition using Parsing Skeleton}, 
      author={Hangrui Xu and Zhengxian Wu and Chuanrui Zhang and Zhuohong Chen and Zhifang Liu and Peng Jiao and Haoqian Wang},
      journal={arXiv preprint arXiv:2503.12047},
      year={2025}
}

🙏 Acknowledgements

• The codebase is based on OpenGait