arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2004.06711cs.CV
收藏

基于可变形暹罗注意网络的视觉目标跟踪

Deformable Siamese Attention Networks for Visual Object Tracking

Malong Technologies(旷视科技)

Yuechen Yu, Yilei Xiong, Weilin Huang, Matthew R. Scott

摘要:

基于Siamese的跟踪器在视觉对象跟踪方面取得了优异的性能。然而,目标模板不在线更新,并且目标模板和搜索图像的特征在Siamese架构中独立计算。在本文中,我们提出了变形暹罗注意力网络,简称为SiamAttn,通过引入一个新的暹罗注意力机制,计算变形的自我注意力和交叉注意力。自我注意通过空间注意学习强上下文信息,并通过通道注意选择性地强调相互依赖的通道特征。交叉关注能够聚合目标模板和搜索图像之间的丰富的上下文相互依赖性,从而提供自适应地更新目标模板的隐式方式。此外,我们还设计了一个区域细化模块,用于计算注意力特征之间的深度相关性,以实现更准确的跟踪。我们在六个基准上进行了实验,其中我们的方法实现了新的最先进的结果,在VOT 2016年和2018年的0.464->0.537和0.415->0.470 EAO上优于强基线SiamRPN++ [24]。我们的代码可在https://github.com/msight-tech/research-siamattn上获得。

英文摘要:

Siamese-based trackers have achieved excellent performance on visual object tracking. However, the target template is not updated online, and the features of the target template and search image are computed independently in a Siamese architecture. In this paper, we propose Deformable Siamese Attention Networks, referred to as SiamAttn, by introducing a new Siamese attention mechanism that computes deformable self-attention and cross-attention. The self attention learns strong context information via spatial attention, and selectively emphasizes interdependent channel-wise features with channel attention. The cross-attention is capable of aggregating rich contextual inter-dependencies between the target template and the search image, providing an implicit manner to adaptively update the target template. In addition, we design a region refinement module that computes depth-wise cross correlations between the attentional features for more accurate tracking. We conduct experiments on six benchmarks, where our method achieves new state of-the-art results, outperforming the strong baseline, SiamRPN++ [24], by 0.464->0.537 and 0.415->0.470 EAO on VOT 2016 and 2018. Our code is available at: https://github.com/msight-tech/research-siamattn.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑