Most deep trackers still follow the guidance of siamese paradigms and use a template that contains only target without any contextual information, which makes it difficult for tracker to cope with large appearance changes, rapid movement, attraction from similar objects. To alleviate above problem, we propose long-term context attention (LCA) module can perform extensive information fusion on i...

Abstract With the continuous development of deep learning in multi-target tracking, use convolutional neural network for feature extraction has replaced traditional method, but accuracy target tracking needs to be improved. In order further improve a new algorithm based on RFB is proposed this paper. The mainly divided into three parts: detection, and tracking. detection part, CenterNet was sel...

In this paper we discuss about a target tracking algorithm based on Mean Shift and Kalman Filter, which is suitable for high speed moving target tracking. The basic Mean Shift algorithm is described in this paper as well. Although basic Mean Shift algorithm can realize target tracking without arguments or searching all the areas effectively, it has shortcomings which can limit its effectiveness...