Revisiting AP Loss for Dense Object Detection: Adaptive Ranking Pair Selection

TL;DR

This paper analyzes the impact of pair selection in AP loss for dense object detection, proposing an adaptive pairwise error loss and a clustering-based method to improve ranking accuracy, leading to better detection performance.

Contribution

It introduces a novel adaptive pairwise error loss and a clustering-based ranking pair selection strategy for AP loss in object detection.

Findings

Proposed APE loss improves detection accuracy on MSCOCO.

Clustering-based pair selection enhances ranking quality.

Experiments show superiority over existing classification and ranking losses.

Abstract

Average precision (AP) loss has recently shown promising performance on the dense object detection task. However,a deep understanding of how AP loss affects the detector from a pairwise ranking perspective has not yet been developed.In this work, we revisit the average precision (AP)loss and reveal that the crucial element is that of selecting the ranking pairs between positive and negative samples.Based on this observation, we propose two strategies to improve the AP loss. The first of these is a novel Adaptive Pairwise Error (APE) loss that focusing on ranking pairs in both positive and negative samples. Moreover,we select more accurate ranking pairs by exploiting the normalized ranking scores and localization scores with a clustering algorithm. Experiments conducted on the MSCOCO dataset support our analysis and demonstrate the superiority of our proposed method compared with current classification and ranking loss. The code is available at https://github.com/Xudangliatiger/APE-Loss.