BDD Dataset is a video dataset that contains 100,000 high-definition video sequences with a total length of more than 1,100 hours, covering different times, weather conditions, and driving scenarios. The video sequences also include GPS locations, IMU data, and timestamps.

Among them, road object detection is to annotate 2D bounding boxes on 100,000 pictures of buses, traffic lights, traffic signs, people, bicycles, trucks, motorcycles, cars, trains and passengers; instance segmentation is used to explore more than 10,000 images with pixel-level and rich instance-level annotations; the engine area is to learn complex drivable decisions from 100,000 pictures; lane marking is a variety of lane annotations on 100,000 driving guide pictures.

This dataset was released by the University of California, Berkeley in 2018, and the related paper is "BDD100K: A Diverse Driving Video Database with Scalable Annotation Tooling".