Channel Positive and Negative Feedback Network for Target Tracking

Aiming at alleviate the detrimental effect of similar object interferences and target state changes in SiamRPN tracker, a Channel Positive Negative Feedback Network (CPFN) is proposed, which Gaussian score map generated by feature channels selected kernel, combined with classification branches SiamRPN. In this way, are divided into positive feedback interference channels, these effectively utilized. addition, channel weight update strategy proposed to enhance robustness tracker avoid template pollution caused inadequate update. Extensive experiments on tracking benchmarks including VOT2016, VOT2018, VOT2019, OTB100, UAV123, LaSOT GOT-10k show that CPFN outperforms state-of-the-art methods based small backbone network terms accuracy achieves high-speed tracking.