Package Face extends its Lerobot platform with training data for autonomous machines
Last year, Hugging Face, the AI Dev platform, launched Lerobot, a collection of open AI models, data sets and tools to help create real world robotics systems. On Tuesday, the face of the hugs teamed up with the startup of IA Yaak to enlarge Lerobot with a training set for robots and cars which can navigate in environments, such as the streets of the city, independently.
The new set, Called learning to drive (L2D)is on a petact in size and contains sensor data that has been installed on cars in German driving schools. L2D captures the data from the camera, GPS and the “vehicle dynamics” of driving instructors and students sailing streets with construction areas, intersections, highways, etc.
There are a number of autonomous training courses open in companies such as Waymo and AI of alphabet. But many of them focus on planning tasks such as detection and monitoring of objects, which require high quality annotations, according to L2D creators – which makes them difficult to evolve.
On the other hand, L2D is designed to support the development of “end -to -end” learning, according to its creators, which helps predict actions (for example, when a pedestrian can cross the street) directly from sensor entries (for example, camera images)
“The AI community can now build autonomous models from start to finish,” wrote a member of the blog team, co-founder of Yaak, the co-founder of Yaak, Harsimrat Sandhawalia and Remi Cadene, member of the AI team for robotics in Hugging Face, in a blog article. “L2D aims to be the largest open-source autonomous data set which allows the AI community of unique and diversified” episodes “for the formation of end-to-end intelligence.”
The embraced face and Yaak plan to carry out “closed loop” tests of the real world of the models formed using L2D and Lerobot this summer, deployed on a vehicle with a safety driver. Companies call the AI community to subject models and tasks on which they would like the models to be assessed, such as navigating roundabouts and parking spaces.
