Attentional Graph Neural Network for 3D Multi-Object Tracking
Xingdi Liu, Ye Liu · 2023
3D Multi-Object Tracking technology is significantly important in many applications such as autonomous driving, robotics and visual surveillance. In recent years, Graph Neural Networks has been introduced into MOT and impressive progress has been made. However, most methods do not take into account the differences in importance of nodes within the neighborhood when updating node features, leading to a lack of description capabilities for local details. In this paper, while modeling the temporal-spatial relationship between objects, we introduce graph attention into GNN network in the message passing module to better aggregate intra-frame node information. Extensive experiments show that our method achieves competitive results on the nuScenes and KITTI datasets, which demonstrate the effectiveness of our method.