Dataset / arXiv / Nov 30, 2023
Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives
Ego-Exo4D connects first-person and third-person observations of skilled activity to study how viewpoint changes affect action and procedural understanding.
Ego-Exo4D pairs wearable and external viewpoints around skilled human activity, creating a richer view of actions than either perspective provides alone.
For capture-system teams, the project illustrates how egocentric footage can be synchronized with broader scene cameras for training, review, and activity understanding.
Open the arXiv source for the dataset protocol, benchmark definitions, full author list, and reported results.
egocentric visionmultiview videoskilled activitydataset
Open original source