NeurIPS 2020

Beta R-CNN: Looking into Pedestrian Detection from Another Perspective

Meta Review

The paper proposes to use a novel 2D beta distribution to model the full-body box and visible box to deal with the pedestrian detection problem in crowded scenes, coping with occlusions. Then authors propose Beta RCNN and BetaNMS. Results are positive. Reviewers noted that although the paper is interesting the results , in terms of comparison and ablation study are fair enough to demonstrate the effectiveness of the proposed method compared to the existing methods. As well the paper is clear but not good enough. After rebuttal one reviewer made higher his/her rate and and in general all reviewers declared to be satisfied by rebuttal. Thus AC agrees with the acceptance of the paper.