An improved Siamese network for object tracking with graph network and attention mechanism

Object tracking is an important and challenging task in computer vision. However, most tracking algorithms are susceptible to challenges such as appearance variation and complex background. To improve the tracking performance, a Siamese network for object tracking with graph convolutional network and attention mechanism is proposed. Graph convolutional network can aggregate feature information from neighboring nodes, and the attention mechanism can improve the weight of critical features of the target in the tracker. Extensive experimental results on the Object Tracking Benchmark (OTB) datasets validate the advantages of the proposed method.