文献类型：期刊文献

中文题名：面向深度学习目标检测模型训练不平衡研究

英文题名：Research on Imbalanced Training of Deep Learning Target Detection Model

作者：贺宇哲[1];何宁[2];张人[1];梁煜博[1];刘晓晓[2]

第一作者：贺宇哲

机构：[1]北京联合大学北京市信息服务工程重点实验室,北京100101;[2]北京联合大学智慧城市学院,北京100101

第一机构：北京联合大学北京市信息服务工程重点实验室

年份：2022

卷号：58

期号：5

起止页码：172-178

中文期刊名：计算机工程与应用

外文期刊名：Computer Engineering and Applications

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD_E2021_2022】；

基金：国家自然科学基金(61872042,61572077);北京市教委科技重点项目(KZ201911417048);北京联合大学人才强校优选计划(BPHR2020AZ01,BPHR2020EZ01);“十三五”时期北京市属高校高水平教授队伍建设支持计划(CIT&TCD201704069);北京联合大学研究生科研创新资助项目(YZ2020K001)。

语种：中文

中文关键词：目标检测;深度学习;不平衡问题;Faster R-CNN

外文关键词：target detection;deep learning;imbalance problem;Faster R-CNN

摘要：目标检测作为计算机视觉的任务之一已经成为研究热点问题。目前,基于深度学习的目标检测算法层出不穷,但大多数情况下学者只关心它们的模型架构,而忽视了其训练过程。目标检测网络在训练过程中会存在明显的不平衡问题,导致模型检测性能降低,不能达到预期的最佳效果。不平衡问题主要包括两个层次,分别是特征图层次和目标函数层次。为了能够充分发挥目标检测模型架构的潜力,实现更好的训练过程,提出利用Balanced Feature Pyramid和Balanced L_(1) Loss两个模块,同时将它们加入到基于ResNet-50-FPN的Faster R-CNN中,目的是解决Faster R-CNN模型在训练过程中存在的特征图层次和目标函数层次的不平衡问题。通过在MSCOCO数据集上验证,实验结果表明平衡后的模型可达到AP是38.5%的结果,比原Faster R-CNN目标检测模型提高了1.1个百分点。
Target detection as one of computer vision tasks has become a hot issue.At present,target detection algorithms depends on deep learning emerge in endlessly,but in most cases,scholars only care about their model architecture and ignore its training process.The target detection network will have obvious imbalance problems during the training process,which will reduce the performance of model detection and fail to achieve the expected best effect.The imbalance problem mainly includes two levels,namely the feature maps level and the objective function level.In order to fully utilize the potential of the target detection model architecture and achieve a better training process,Balanced Feature Pyramid and Balanced L_(1) Loss modules are proposed to use,and added to the Faster R-CNN based on ResNet-50-FPN,and the purpose is to solve the imbalance between the feature map level and the objective function level in the training process of Faster R-CNN model.Through verification on the MSCOCO dataset,experimental results show that the balanced model can reach a result of 38.5%AP,which is 1.1 percentage points higher than original Faster R-CNN target detection model.