PIE stands for Pedestrian Intention Estimation, which is a dataset for studying pedestrian behavior in traffic. The dataset contains more than 6 hours of video footage of typical traffic scenes, recorded by on-board cameras. It also provides accurate vehicle information (speed, direction, and GPS coordinates) from OBD sensors and is synchronized with the video footage.

Rich spatial and behavioral annotations are available for pedestrians and vehicles that may interact with the ego vehicle, as well as relevant elements of infrastructure (traffic lights, signs, and zebra crossings). The dataset contains more than 300,000 annotated video frames and 1,842 pedestrian samples, making it the largest publicly available dataset for studying pedestrian behavior in traffic.