Researchers at Meta and 15 universities have created a dataset to train AI systems to complete complex tasks. It includes over 1,400 hours of footage of experts completing tasks, such as cooking, playing music, and making repairs, from both a first-person point of view and a third-person point of view. It also features the participants’ narrations of their actions, third-person narrations and critiques of actions, and data on participants’ movements, such as eye gaze and head poses. Researchers can use the dataset to train AI systems for virtual coaching programs, robots, and more.
↧