Researchers at New York University and Meta have created a dataset to train AI systems to direct generalist household robots. The dataset contains 13 hours of footage of people using a gripper stick to complete household tasks, such as picking up clothing or closing cabinets, in 216 settings in 22 homes. It improves upon past household robot datasets by using first-person footage and common household settings instead of laboratories.
Image credit: Flickr user Michael Dain
