NeurIPS 2020

Online Decision Based Visual Tracking via Reinforcement Learning


Meta Review

The initial scores were 3478. the main concerns were: 1) should combine the framework with SOTA trackers; 2) insufficient experiments on larger datasets and current methods; 3) lack fo deeper ablation study; 4) comparison with hand-designed rules; 5) incremental novelty. In the response, authors provide experiment results fusing SOTA trackers on the larger datasets, compared with SOTA trackers, showing improved performance. Authors also provide ablation study using hand-designed rules. During discussion, R2 was satisfied with the new comparisons with SOTA and larger datasets, but was not convinced that the fusion method was useful since there was no ablation study comparing only fusion methods (while keeping trackers the same). R3 was mostly satisfied with the response, but novelty concern was not addressed fully. After the discussion, R2 and R3 upgraded to 4 and 6, leaving the final score as 4678. 3 out fo 4 reviewers are positive on the paper, while R2 was mainly concerned about the ablation study comparing only fusion methods, while keeping baseline trackers the same. To be fair to the authors, the AC notes that in the original review, R2 did not state specifically what type of deeper ablation study was needed, thus making it difficult for the authors to address. Nonetheless, the method is compelling and improves SOTA. Thus, the AC recommends accept. Authors should update the paper according to the reviews/responses, including the missing ablation study