Adaptive NMS: Refining Pedestrian Detection in a Crowd
This addresses the problem of accurate pedestrian detection in dense crowds for applications like autonomous driving and surveillance, representing a strong incremental improvement.
The paper tackled pedestrian detection in crowded scenes by proposing Adaptive NMS, a novel algorithm that uses dynamic suppression thresholds based on target density, achieving state-of-the-art results on CityPersons and CrowdHuman benchmarks.
Pedestrian detection in a crowd is a very challenging issue. This paper addresses this problem by a novel Non-Maximum Suppression (NMS) algorithm to better refine the bounding boxes given by detectors. The contributions are threefold: (1) we propose adaptive-NMS, which applies a dynamic suppression threshold to an instance, according to the target density; (2) we design an efficient subnetwork to learn density scores, which can be conveniently embedded into both the single-stage and two-stage detectors; and (3) we achieve state of the art results on the CityPersons and CrowdHuman benchmarks.