Cross-Correlation Fusion Graph Convolution-Based Object Tracking

Most popular graph attention networks treat pixels of a feature map as individual nodes, which makes the embedding extracted by convolution lack integrity object. Moreover, matching between template and search using only part-level information usually causes tracking errors, especially in occlusion similarity situations. To address these problems, we propose novel end-to-end framework that has high symmetry, combining traditional cross-correlation operations directly. By utilizing operations, effectively compensate for dispersion nodes enhance representation features. Additionally, our fusion model performs both part-to-part global matching, allowing more accurate regions. Furthermore, optimize branches to achieve better single-object results, particularly scenarios. The flexibility comprehensiveness have brought significant performance improvements framework. Extensive experiments on three challenging public datasets (LaSOT, GOT-10k, VOT2016) show tracker outperforms other state-of-the-art trackers.