随着这两年 GNN 的发展，其对于关系的建模特性也被引入了多目标跟踪领域，这次我通过对这两年基于 GNN 的 MOT 算法的介绍来分析其特点。

论文标题： Graph Neural Based End-to-end Data Association Framework for Online Multiple-Object Tracking 论文来源：CVPR 2020 论文链接：https://arxiv.org/abs/1907.05315 代码链接：https://github.com/peizhaoli05/EDA_GNN

论文标题：Deep association: End-to-end graph-based learning for multiple object tracking with conv-graph neural network 论文来源：ICMR 2019 论文链接：https://dl.acm.org/doi/pdf/10.1145/3323873.3325010

论文标题：Graph Networks for Multiple Object Tracking 论文来源：WACV 2020 论文链接：http://openaccess.thecvf.com/content_WACV_2020/papers/Li_Graph_Networks_for_Multiple_Object_Tracking_WACV_2020_paper.pdf 代码链接：https://github.com/yinizhizhu/GNMOT

论文标题：Learning a Neural Solver for Multiple Object Tracking 论文来源：CVPR 2020 论文链接：https://arxiv.org/abs/1912.07515 代码链接：https://github.com/selflein/GraphNN-Multi-Object-Tracking

论文标题：Graph Neural Network for 3D Multi-Object Tracking with Multi-Feature Learning 论文来源：CVPR 2020 论文链接：https://arxiv.org/abs/2006.07327 代码链接：https://github.com/xinshuoweng/GNN3DMOT

论文标题：Joint 3D Tracking and Forecasting with Graph Neural Network and Diversity Sampling 论文链接：https://arxiv.org/abs/2003.07847 代码链接：https://github.com/xinshuoweng/GNNTrkForecast

内容中包含的图片若涉及版权问题，请及时与我们联系删除