Label assignment has been widely studied in general object detection because of its great impact on detectors’ performance . In this paper, we propose a simple yet effective assigning strategy called Loss-aware LabelAssignment (LLA) to boost the performance of pedestrian detectors in crowdscenarios . LLA first calculates classification (cls) and regression (reg)losses) between each anchor and ground-truth (GT) pair . Anchors with lower joint loss usually contain richer semanticinformation and thus can better represent their corresponding GT boxes . Experimented on CrowdHuman and CityPersons show that such a simple labelassigning strategy can boost MR by 9.53% and 5.47% on two famous one-stagedetectors – RetinaNet and FCOS, respectively, demonstrating the effectivenessof LLA LLA. LLA

Author(s) : Zheng Ge, Jianfeng Wang, Xin Huang, Songtao Liu, Osamu Yoshie

Links : PDF - Abstract

Code :


Keywords : lla - loss - aware - strategy - boost -

Leave a Reply

Your email address will not be published. Required fields are marked *