Skip to main content
Banner

Closed-Loop Full-Stack Benchmark For VLMs in Autonomous Driving

Demo Video

Dynamic Annotated VQA

Dynamic Annotated VQA

Our rule-based expert DriveCommenter is capable of generating abundant VQAs both from static datasets and CARLA simulator runtime.

Adapting Vision Language Models to CARLA

Adapting Vision Language Models to CARLA

Both VLMs and CARLA have complicated dependencies. Bench2Drive-VL adapts VLM to CARLA simulator via various configurable approaches.

Community Support

Community Support

Our community provides technical supports about CARLA and Bench2Drive. Docker file included!