TACO is a comprehensive dataset for studying bimanual hand-object interactions, supporting generalizable research in action recognition, motion forecasting, and grasp synthesis.
TACO dataset provides insights into bimanual hand-object interactions for generalizable research.