Consistent GT-Proposal Assignment for Challenging Pedestrian Detection

Accurate pedestrian classification and localization has garnered significant attention due to their extensive applications in various multimedia applications such as security monitoring, autonomous driving, and more. We have observed that the commonly employed Intersection over Union (IoU) metric in many pedestrian detectors is susceptible to an inconsistent GT-Proposal assignment issue. This issue arises when spatially adjacent proposals, which have highly similar features, are assigned to distinct ground-truth boxes, leading to confusion during the training process and an increased number of false positives during inference. To address this challenge, our work presents a novel algorithm named Directional Assignment Strategy (DAS). Firstly, in conjunction with depth distribution, our approach transforms the assignment metric from a two-dimensional (2D) view into a three-dimensional (3D) space, enabling the optimization of the regression head under the constraint of depth direction. Secondly, in contrast to the conventional IoU-based one-to-one assignment of one proposal to one ground-truth box, our method aims to establish a more reasoned matching between sets of proposals and ground-truth boxes. By doing so, the detector is less reliant on the setting of a specific threshold. Leveraging this strategy as a plug-in module within state-of-the-art pedestrian detectors, we demonstrate a notable improvement in performance.

I. Introduction

Pedestrian detection serves as a fundamental building block for various multimedia applications such as surveillance systems and autonomous driving. Driven by the success of general object detection, many current pedestrian detectors [1], [2], [3], [4], [5], built on the basic practice such as Faster R-CNN [6] and SSD [7], have shown good performance on most pedestrian detection scenes. However, with the increasing demand for accuracy in autonomous driving, their performance in handling challenging pedestrian detection scenarios, particularly cases involving occlusion or significant scale variation, is currently unsatisfactory.