In this paper, we formulate the multi-tracker tracking problem as a decision-making task and train an expert by the deep reinforcement learning (DRL) to select the best tracker. Specifically, the expert takes the response map of the tracker as input and outputs a scalar to indicate the reliability of the tracker. With the DRL, the expert can make full use of complementary information among base trackers. Furthermore, under the guidance of the deep expert, base trackers update themselves adaptively to capture the changes of object appearance and prevent corruption. The experimental results on public tracking benchmarks demonstrate that the proposed method outperforms the state-of-the-art methods.
展开▼
