MoVi Dataset

A large multi-purpose human motion and video dataset with synchronized pose, body meshes, and video recordings

Overview

MoVi is a comprehensive human motion and video dataset that provides synchronized pose, body meshes, and video recordings. The dataset enables research in human motion analysis, pose estimation, action recognition, and motion synthesis.

Key Features

  • Multi-modal data with synchronized motion capture, video, and IMU recordings
  • 90 actors (60 female, 30 male) performing over 20 different actions
  • Multiple camera viewpoints including handheld perspective
  • Full-body pose estimation and shape deformation annotations
  • Data captured under different clothing conditions
MoVi Dataset Examples
Sample motion sequences from the MoVi dataset showing synchronized multi-modal data capture

Dataset Statistics

Metric Value
Total Actors 90 (60F, 30M)
Actions 20+ everyday and sports movements
Motion Capture Data 9 hours
Video Data 17 hours (4 viewpoints)
IMU Data 6.6 hours
Capture Rounds 5 different hardware systems

Applications

Research Applications

  • Human pose estimation and tracking from multiple viewpoints
  • Motion prediction and forecasting
  • Action recognition and classification
  • Gait analysis and biomechanical studies

Practical Applications

  • Character animation for games and films
  • Sports performance analysis
  • Healthcare and rehabilitation monitoring
  • Human-computer interaction systems

Technical Details

Data Modalities

  1. Optical Motion Capture: High-precision 3D joint positions and orientations
  2. Video Recordings: Multi-view RGB video at various resolutions
  3. IMU Sensors: Inertial measurement data for wearable applications
  4. Body Meshes: Full-body shape deformations using SMPL model

Hardware Systems

The dataset was captured using professional optical motion capture systems, multiple synchronized video cameras, and wireless IMU sensor networks under controlled lighting conditions.

Impact

MoVi was the first dataset to provide synchronized multi-modal human motion data, making it a valuable resource for the computer vision and graphics research communities. The dataset has been widely adopted for various research applications and continues to support advances in human motion understanding.

Resources

Publication

Published in PLOS ONE - Read the full paper

Code & Tools

Complete toolbox and tutorials available on GitHub

Official Website

Visit the main MoVi website for dataset download and documentation

Dataset Access

The dataset is freely available for research purposes. Please visit the official website for access instructions and licensing information.

Citation

If you use the MoVi dataset in your research, please cite:

@article{ghorbani2021movi,
  title={MoVi: A large multi-purpose human motion and video dataset},
  author={Ghorbani, Saeed and Mahdaviani, Kimia and Thaler, Anne and Kording, Konrad and Cook, Douglas James and Blohm, Gunnar and Troje, Nikolaus F},
  journal={Plos one},
  volume={16},
  number={6},
  pages={e0253157},
  year={2021},
  publisher={Public Library of Science San Francisco, CA USA}
}