Skip to main content

HOT3D Dataset

HOT3D is a new benchmark dataset for vision-based understanding of 3D hand-object interactions. This dataset contains over 800 minutes of egocentric recordings, with 33 diverse hand-held objects, capturing over one million multi-view frames of hand-object interactions.

The dataset contains:

  • Synchronized multi-view egocentric videos from Project Aria glasses and Quest 3 VR headset
  • High-quality 3D pose annotations of hands and objects
  • 3D object models with PBR materials
  • 2D bounding boxes
  • Eye Gaze MPS data (Aria only)
  • Semi-Dense Point Cloud MPS data (Aria only)

HOT3D uses its own specific downloader, available in the HOT3D GitHub repository, enabling you to download Quest3, Aria and object models data.

Getting Started

HOT3D Research Challenges

HOT3D data is used in the following research challenges: