<ul><li>Understanding human motion is crucial for accurate pedestrian trajectory prediction.</li><li>This work proposes a self-supervised pedestrian trajectory prediction framework that explicitly models position, velocity, and acceleration.</li><li>The model leverages velocity and acceleration information to enhance position prediction through feature injection and a self-supervised motion consistency mechanism.</li><li>Experiments on the ETH-UCY and Stanford Drone datasets show that the proposed method achieves state-of-the-art performance.</li></ul>

Learning Velocity and Acceleration: Self-Supervised Motion Consistency for Pedestrian Trajectory Prediction

Discover more