Project Page | Paper | Code
Official CARLA dataset accompanies our paper LEAD: Minimizing Learner–Expert Asymmetry in End-to-End Driving.
We release the complete pipeline required to achieve state-of-the-art closed-loop performance on the Bench2Drive benchmark. Built around the CARLA simulator, the stack features a data-centric design with:
Extensive visualization suite and runtime type validation for easier… See the full description on the dataset page:
https://huggingface.co/datasets/ln2697/lead.