Core Concepts
Efficient NMS algorithms, QSINMS and BOENMS, are proposed based on a graph theory analysis of the intrinsic structure of NMS, achieving significant speedups without compromising detection accuracy.
Abstract
The paper presents a comprehensive analysis of the NonMaximum Suppression (NMS) algorithm from a graph theory perspective. It reveals the intrinsic structure of NMS, where the set of bounding boxes can be mapped to a directed acyclic graph (DAG).
Based on this analysis, the authors propose two optimization methods:

QSINMS (Quicksort Induced NMS):
 Exploits the independence of weakly connected components (WCCs) in the NMSinduced graph.
 Uses a divideandconquer approach inspired by quicksort to efficiently solve the problem.
 The extended version, eQSINMS, achieves an optimal time complexity of O(n log n).

BOENMS (Boxes Outside Excluded NMS):
 Leverages the locality of box distributions, where most WCCs are small.
 Avoids computing IOUs for boxes that cannot have suppression relationships based on geometric analysis.
 Achieves constantlevel optimization without compromising detection accuracy.
The authors also introduce NMSBench, the first endtoend benchmark for rapid validation of NMS algorithms. It decouples model inference and postprocessing, enabling efficient evaluation of NMS methods.
Experiments on MS COCO 2017 and Open Images V7 datasets demonstrate the superiority of the proposed methods. QSINMS and eQSINMS achieve up to 16.9x and 8.9x speedups, respectively, compared to original NMS, with negligible accuracy loss. BOENMS provides up to 9.1x speedup without any accuracy compromise.
Stats
The paper reports the following key metrics:
Average latency of different NMS methods on MS COCO 2017 and Open Images V7 datasets.
Mean Average Precision (mAP) of different NMS methods on the same datasets.
Quotes
"We present the first comprehensive analysis of the NMS algorithm from a graph theory perspective, uncovering the intrinsic structure of NMS."
"We propose two efficient NMS algorithms based on the properties of the NMSinduced graph."
"We introduce NMSBench, the first endtoend benchmark for rapid NMS validation."