论文标题
利用地图信息进行自我监督的学习预测
Exploiting map information for self-supervised learning in motion forecasting
论文作者
论文摘要
受到有关应用自我监督学习（SSL）的最新发展的启发，我们设计了轨迹预测的辅助任务，该任务利用了仅MAP的信息，例如图形连接性，以改善地图理解和概括。我们通过两个框架（多任务处理和预处理）应用了这项辅助任务。在任何一个框架上，我们都会观察到基线在诸如$ \ mathrm {minfde} _6 $（多达20.3％）和$ \ mathrm {sillrate} _6 $（多达33.3％）的指标中的显着改善，以及不同培训配置表明的地图功能的丰富理解。在用于实验的所有三个数据集中，获得的结果都是一致的：argoverse，互动和努斯曲霉。我们还将新验证的模型的结果提交给互动挑战，并获得相对于$ \ mathrm {minfde} _6 $和$ \ mathrm {minade} _6 $的$ \ textit {1st} $ lote。
Inspired by recent developments regarding the application of self-supervised learning (SSL), we devise an auxiliary task for trajectory prediction that takes advantage of map-only information such as graph connectivity with the intent of improving map comprehension and generalization. We apply this auxiliary task through two frameworks - multitasking and pretraining. In either framework we observe significant improvement of our baseline in metrics such as $\mathrm{minFDE}_6$ (as much as 20.3%) and $\mathrm{MissRate}_6$ (as much as 33.3%), as well as a richer comprehension of map features demonstrated by different training configurations. The results obtained were consistent in all three data sets used for experiments: Argoverse, Interaction and NuScenes. We also submit our new pretrained model's results to the Interaction challenge and achieve $\textit{1st}$ place with respect to $\mathrm{minFDE}_6$ and $\mathrm{minADE}_6$.
