TL;DR本文提出了基于局部相关性模块的密集对应和可学习相关算子来增强模型的判别能力和对时间上下文的捕捉能力,从而在多目标跟踪方面取得了最先进的效果,并在 MOT17 数据集上实现了 76.5% 的 MOTA 和 73.6% 的 IDF1。
Abstract
Recent works have shown that convolutional networks have substantially
improved the performance of multiple object tracking by simultaneously learning
detection and appearance features. However, due to the local