Yijie's research area is pedestrian crossing intention prediction. The current mainstream approach is to use the features obtained from the intermediate algorithm to make the next prediction. We hope to further optimize the datasets and models to implement a vision-based prediction algorithm.