Spatio-Temporal Human Action Recognition and Localization
A spatio-temporal action recognition project using contextual relationships between actions, people, and surrounding objects.
Project Snapshot
| Signal | Details |
|---|---|
| Timeline | Jul 2022 to Nov 2022 |
| Organization | XPENG Robotics |
| Focus | Spatio-Temporal Human Action Recognition |
| Stack | Custom implementation and project-specific tooling |
Methods Explored
- Studied ACAR-Net, AIA, and long-term feature bank approaches.
- Manually labeled custom 30 FPS video data for ACAR-Net training.
- Trained ACAR-Net on custom action recognition data.
Action Categories
| Category | Examples |
|---|---|
| Individual action | Run, walk, jump, dance |
| Human-object interaction | Carry, hold, touch an object |
| Human-human interaction | Hug, assist, approach |
| Human-object-human interaction | Give, serve, take an object |