Visual Object Tracking via Graph Convolutional Representation
Zhengzheng Tu, Ajian Zhou, Bo Jiang, Bin Luo · 2019
CNN-based trackers are easily interfered by insufficient feature learning, causing model drift. In recent years, graph convolutional networks (GCNs) have been widely used for the representation of graph data in the fields of machine learning and computer vision. In our work, we employ a GCN module to learn structural features for visual tracking. First, we utilize a dual path network to extract heterogeneous features. Then, we adopt a GCN module to construct features to have structured information. Finally, we connect all the features and use the attention mechanism to adaptively select features. Extensive experiments on two benchmark datasets validate the effectiveness of our approach.