Cross-Covariate Gait Recognition: A Benchmark

Welcome to the official repository for the paper "Cross-Covariate Gait Recognition: A Benchmark," which has been accepted to AAAI 2024. In this work, we introduce a novel dataset and parsing-based gait recognition method. Below you'll find details about the paper, dataset, and accompanying resources.

Paper Links

• AAAI 2024 Version: Download paper (AAAI)
• ArXiv Preprint: Download preprint (ArXiv)

Dataset Visualization

Here are some visualizations that depict the characteristics of the Cross-Covariate Gait Recognition (CCGR) dataset in comparison with other existing datasets, as well as the distribution of covariates, viewpoints, and data modalities within CCGR.

CCGR Dataset vs. Other Datasets

Covariates in the CCGR Dataset

Views Distribution in the CCGR Dataset

Data Modalities in the CCGR Dataset

CCGR Dataset Download Guide

Derived data (Silhouette, Parsing, Pose)

We are pleased to offer the derived data for use in your research projects. To ensure proper usage, we kindly ask that you adhere to the following terms:

1. Research Purposes Only: The dataset is strictly for academic and non-commercial research purposes only.
2. No Modification or Redistribution: Modifying or distributing the dataset in any form is not permitted.
3. Internal Use: The dataset is to be used within your organization or research group and is not to be shared outside of this group.

By downloading the dataset, you indicate your acceptance of these terms. We have facilitated easy access to the data via the links provided below, so no additional download request is necessary!

Dataset Links

You are invited to download the CCGR dataset from the links below. Before proceeding with the download, please ensure you have agreed to the above terms.

• Baidu Netdisk: Download Dataset (Code:ngcw)
• OneDrive: Download Dataset

We wish to support a collaborative and respectful research community and we believe that adherence to these terms will be mutually beneficial.

Raw data (RGB)

We would like to inform you that the Raw data is scheduled to be released on March 1, 2024, because we need more time to upload the data (the raw data is bigger). Details regarding access and download instructions will be announced as the date approaches. Please stay tuned to our official communication channels for further updates.

Thank you for your attention and cooperation. We sincerely hope the dataset aids you in your research endeavors.

Ethical Statement

All subjects are openly recruited and participate voluntarily, and they need to read and acknowledge the collection protocol by signature and fingerprint. In return, we pay each subject a fee for data copyright.

Testing Code on OpenGait Platform

Specific adjustments are necessary to utilize the OpenGait framework testing the CCGR dataset effectively. For detailed instructions on how to adapt OpenGait, please refer to the guidelines provided at the beginning of the CCGR_EVA.py file. These modifications are designed to ensure seamless integration of the CCGR dataset with the existing capabilities of OpenGait. Additionally, we have included some yaml files within this document and some trained weights for silhouette. Baidu Netdisk (Code:4o92); OneDrive.

• OpenGait Platform

Cite Us

If you find our dataset or paper useful for your research, please consider citing:

@inproceedings{Zou2024CCGR,
  title={Cross-Covariate Gait Recognition: A Benchmark},
  author={Zou, Shinan and Fan, Chao and Xiong, Jianbo and Shen, Chuanfu and Yu, Shiqi and Tang, Jin},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2024}
}