TraceVLA visual trace prompting dataset

Name: TraceVLA visual trace prompting dataset
Creator: umd-huang-lab (GitHub org)
Keywords: research-lab

TraceVLA visual trace prompting dataset is a language and reasoning annotation robotics dataset maintained by umd-huang-lab (GitHub org). It contains 150,000 episodes of demonstration data. Primary hardware platform: robot manipulation trajectories; evaluated on WidowX and SimplerEnv. Access is not stated under not stated.

Dataset specifications
Year	2024
Episodes	150,000
Embodiments	robot manipulation trajectories; evaluated on WidowX and SimplerEnv
Task categories	manipulation
License	not stated
Access	open
Maintainer	umd-huang-lab (GitHub org)