Researchers at the University of California, Berkeley, Stanford University, Google, and Carnegie Mellon University have created a dataset to improve robotic learnings. The dataset contains over 60,000 videos of a robotic arm performing tasks, such as opening drawers, stacking blocks, and folding fabrics, in 24 different environments, as well as natural language instructions for each task. The team teleoperated the arm in over 50,000 of the videos and used an AI system to direct the arm in over 9,700 of the videos.
↧