Hugging Face, the AI development platform, joined forces with AI startup Yaak to expand its robotics initiative, LeRobot.
Recently, the companies introduced Learning to Drive (L2D), a massive training set exceeding one petabyte, designed to advance autonomous navigation in robots and cars. L2D features sensor data from vehicles at German driving schools, capturing real-world environments like highways, construction zones, and intersections. Unlike existing self-driving datasets, L2D emphasizes “end-to-end” learning, predicting actions directly from sensor inputs.
Hugging Face and Yaak plan real-world testing of models trained with L2D and LeRobot this summer, inviting the AI community to contribute models for evaluation on tasks like roundabout navigation and parking. This collaboration aims to empower innovation in self-driving technology.