Deep Learning Based Pedestrian Trajectory Prediction Considering Location Relationship between Pedestrians
In-Kyu Choi, Hyok Jae Song, Ji‐Sang Yoo · 2019
Pedestrian trajectory prediction is a challenging task because of the complex nature of humans. We propose to predict displacement between neighboring frames for each pedestrian sequentially. Specifically, we use an LSTM to model motion information for all pedestrians and use a mlp to map the location of each pedestrian to a high dimensional feature space where the inner product between features is used as a measurement for the positional relationship between two pedestrians. Then we weight the motion features of all pedestrians based on their positional relationship to the target for location displacement prediction. Experiments on publicly available datasets validate the effectiveness of our method for trajectory prediction.