hot3d - Multi-View 3D Hand & Object Tracking
Multi-view egocentric 3D hand and object tracking system for Aria/Quest smart glasses
Tags
Updated: 2026-03-20Capabilities
Typical Inputs
Typical Outputs
What this skill does
- track 3D hand pose
- estimate object pose
- detect interactions
- visualize 3D overlays
- project 3D skeleton
- render hand meshes
- overlay object models
Inputs
- multi-view RGB-D video streams
- Aria/Quest device data
- HOT3D dataset
- camera calibration data
Outputs
- 3D hand joint coordinates
- object 3D positions
- object rotation matrices
- 3D overlay visualizations
- hand mesh models
- trajectory files
Requirements
- multi-camera setup or Aria/Quest device
- GPU (NVIDIA RTX 3080 or better)
- PyTorch
- PyTorch3D
- Open3D
