HEV-I Trajectory Prediction Video Dataset

HHEV-I stands for Honda Egocentric View-Intersection, which contains 230 driving videos at different intersections in the San Francisco Bay Area, collected by sensors installed on the vehicle (camera, GPS/IMU, and vehicle status signals). This dataset can be used to study the interaction modeling of traffic participants, future target positioning, and learning driving behavior in challenging driving scenarios.

The dataset includes:

• 230 videos with a resolution of 1280 X 640 and a frequency of 10 Hz.
• Target objects: cars, pedestrians, bicycles, motorcycles, trucks, buses, traffic lights, bus stops.
• Ego behaviors: Go straight, accelerate, brake, turn right, turn left, yield (to pedestrians, bicycles, or cars), stop (at a stop sign or traffic light).