10,000 hand-object interaction sequences across 100 object categories featuring affordance-driven hand poses. The dataset includes 3D hand meshes, object models, and contact maps representing functional hand-object interactions.
Use Cases
- Train generative models for hand pose synthesis using the affordance labels and object meshes
- Develop contact-aware hand-object interaction algorithms using the provided contact maps
- Benchmark 3D hand pose estimation in the context of functional object manipulation
Strengths
- 10,000 hand-object interaction sequences
- 100 distinct object categories with corresponding 3D meshes
- Includes affordance-driven hand poses and contact maps
- Presented at ICCV 2023