目标检测强化上下文模型  

Object detection enhanced context model

作　　者：郑晨斌 张勇[1] 胡杭 吴颖睿 黄广靖 ZHENG Chen-bin;ZHANG Yong;HU Hang;WU Ying-rui;HUANG Guang-jing(School of Instrumetation and Optoelectronic Engineering,Beihang University,Beijing 100191,China;Unit 66133 of PLA,Beijing 100144,China;School of Aeronautic Science and Engineering,Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100191 [2]解放军66133部队,北京100144 [3]北京航空航天大学航空科学与工程学院,北京100191

出　　处：《浙江大学学报（工学版）》2020年第3期529-539,共11页Journal of Zhejiang University：Engineering Science

摘　　要：强化上下文模型中的强化上下文模块(ECM)利用双空洞卷积结构,在节省参数量的同时,通过扩大有效感受野来强化浅层上下文信息,并在较少破坏原始SSD网络的基础上灵活作用于网络中浅预测层,形成强化上下文模型网络(ECMNet).当输入图像大小为300×300时,在PASCAL VOC2007测试集上,ECMNet获得的均值平均精度为80.52%,在1080Ti上的速度为73.5帧/s.实验结果表明,ECMNet能有效强化上下文信息,并在参数量、速度和精度上达到较优权衡,优于许多先进的目标检测器.Double-atrous convolution structure was used in enhanced context module(ECM) of the enhanced context model to reduce parameters while expanding effective receptive field to enhance context information of shallow layers, and ECM flexibly acted on middle shallow prediction layers with less damage to original SSD, forming enhanced context model net(ECMNet). Using input image with size of 300×300, ECMNet obtained mean average precision of 80.52% on PASCAL VOC2007 test set, and achieved 73.5 frames per second on 1080 Ti. The experimental results show that ECMNet can effectively enhance context information and achieves a better trade-off in parameter, speed and accuracy, which is superior to many state-of-the-art object detectors.

关 键 词：目标检测 上下文信息 有效感受野 强化上下文模块(ECM) 一阶段目标检测器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]