This egocentric video dataset captures the range of actions that occur in indoor work spaces, performed with the objects those environments contain, for Physical AI model training.
Potential Use Cases
- Workspace Task Policy Training: Demonstrations set in work environments supply the object types and action sequences of desk and tool-based tasks, which domestic recordings do not cover.
- Environment-Conditioned Action Modeling: Holding the setting fixed while the actions vary lets a model learn which behaviours a workspace affords, useful for assistive agents that must anticipate the next step.