ShiqiYu/OpenGait

A flexible and extensible framework for gait recognition. You can focus on designing your own models and comparing with state-of-the-arts easily with the help of OpenGait.

What it solves

OpenGait provides a flexible and extensible framework for gait analysis, specifically focusing on gait recognition (identifying individuals by their walking patterns). It addresses the need for a comprehensive benchmark and a unified platform to implement, test, and compare various gait recognition methods across different data modalities.

How it works

The project implements a wide array of gait recognition models that process different types of input data, including silhouettes, skeleton maps, and LiDAR point clouds. It supports multiple state-of-the-art (SOTA) methods—such as BigGait (using Large Vision Models), SkeletonGait++, and LidarGait++—and provides a standardized pipeline for training and testing. The framework includes support for Distributed Data Parallel (DDP) for efficient training and Auto Mixed Precision (AMP) to optimize performance.

Who it’s for

It is designed for academic researchers and developers working in computer vision, biometric identification, and healthcare (e.g., using gait patterns as biomarkers for scoliosis screening).

Highlights

  • Broad Modality Support: Handles silhouette-based, pose/skeleton-based, and 3D LiDAR-based gait analysis.
  • Extensive Model Zoo: Includes a variety of reproduced SOTA models and new contributions like FoundationGait and BiggerGait.
  • Comprehensive Dataset Integration: Supports numerous datasets including CASIA-B, OUMVLP, GREW, and the million-level GaitLU-1M.
  • Clinical Application: Includes specialized tools like ScoNet for scoliosis screening based on gait patterns.

Related

  • Project
    PetoiCamp/OpenCat-Quadruped-Robot

    An open-source framework for quadruped robots that handles gait coordination and servo control, enabling the development of AI-driven robotic pets and embodied intelligence research.

  • Project
    XiandaGuo/OpenStereo

    A comprehensive benchmark and framework for stereo matching that integrates 17 datasets and multiple state-of-the-art models for depth estimation.

  • Project
    yzfzzz/depth-detect

    A high-performance C++ and TensorRT framework that fuses YOLO object detection with monocular/binocular depth estimation to track objects and estimate their motion states.

  • Project
    talmolab/sleap

    A deep-learning framework for multi-animal pose tracking that uses a human-in-the-loop GUI to rapidly label and quantify animal behavior.

  • Project
    sjtuyinjie/M3DGR

    M3DGR is a multi-sensor SLAM benchmark dataset and resilient framework designed to test and improve the robustness of ground robot localization and mapping in degraded environments.