JAAD stands for Joint Attention in Autonomous Driving, which is a dataset used to study joint attention in the context of autonomous driving. The focus is on the behavior of pedestrians and drivers at intersections and the factors that affect them.

The dataset contains 346 video clips with a frame rate of 30, a duration of 5-10 seconds, and a total of 82,032 frames. The videos are collected by 3 on-board cameras and cover typical scenes of daily urban driving in various weather conditions in North America and Eastern Europe.

The annotation of this dataset has the following characteristics:

• 2,793 pedestrians, 378,643 2D annotation boxes

• Different label boxes: walking, standing, crossing the road, checking traffic conditions

• Pedestrian attributes: age, gender, clothing, movement direction

• Each video has time and weather attributes