Pedestrian Graph: Pedestrian Crossing Prediction Based on 2D Pose Estimation and Graph Convolutional Networks

Pedestrian crossing prediction is a particularly important task for intelligent transportation systems, accurate prediction can guarantee the safety of pedestrians and driving comfort of vehicles. This paper predicts intentions of pedestrians crossing on urban roads based on 2D human pose estimation and Graph Convolutional Network (GCN), achieving the new state-of-the-art in the Joint Attention in Autonomous Driving (JAAD) data set. The major contribution of this work is the development of the 2D pedestrian graph structure and pedestrian graph network to predict whether a pedestrian is going to cross the street. The proposed method obtained an accuracy of 91.94 % in pedestrian crossing prediction.