Small Object Detection via Precise Region-Based Fully Convolutional Networks

作     者：Dengyong Zhang Jiawei Hu Feng Li Xiangling Ding Arun Kumar Sangaiah Victor SSheng 

作者机构：Hunan Provincial Key Laboratory of Intelligent Processing of Big Data on TransportationChangsha University of Science and TechnologyChangsha410114China School of Computer and Communication EngineeringChangsha University of Science and TechnologyChangsha410114China School of Computer Science and EngineeringHunan University of Science and TechnologyXiangtan411004China School of Computing Science and EngineeringVellore Institute of Technology(VIT)Vellore632014India Department of Computer ScienceTexas Tech UniversityLubbock79409TXUSA 

出 版 物：《Computers, Materials & Continua》 (计算机、材料和连续体（英文）)

年 卷 期：2021年第69卷第11期

页      面：1503-1517页

核心收录：

学科分类：0831[工学-生物医学工程（可授工学、理学、医学学位）] 0808[工学-电气工程] 0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 0805[工学-材料科学与工程（可授工学、理学学位）] 0701[理学-数学] 0801[工学-力学（可授工学、理学学位）] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：This project was supported by the National Natural Science Foundation of China under grant U1836208 the Hunan Provincial Natural Science Foundations of China under Grant 2020JJ4626 the Scientific Research Fund of Hunan Provincial Education Department of China under Grant 19B004 the“Double First-class”International Cooperation and Development Scientific Research Project of Changsha University of Science and Technology under Grant 2018IC25 the Young Teacher Growth Plan Project of Changsha University of Science and Technology under Grant 2019QJCZ076 

主　　题：Small object detection precise R-FCN PS-Pr-RoI pooling two-stage detector 

摘      要：In the past several years,remarkable achievements have been made in the field of object *** performance is generally improving,the accuracy of small object detection remains low compared with that of large object *** addition,localization misalignment issues are common for small objects,as seen in GoogLeNets and residual networks(ResNets).To address this problem,we propose an improved region-based fully convolutional network(R-FCN).The presented technique improves detection accuracy and eliminates localization misalignment by replacing positionsensitive region of interest(PS-RoI)pooling with position-sensitive precise region of interest(PS-Pr-RoI)pooling,which avoids coordinate quantization and directly calculates two-order integrals for position-sensitive score maps,thus preventing a loss of spatial precision.A validation experiment was conducted in which the Microsoft common objects in context(MS COCO)training dataset was *** showed an accuracy improvement of 3.7%for object detection tasks and an increase of 6.0%for small objects.