首页 | 官方网站   微博 | 高级检索  
     

DF-SSD:一种基于反卷积和特征融合的单阶段小目标检测算法
引用本文:王良玮,陈梅,李晖,李焕军,施若,戴震宇. DF-SSD:一种基于反卷积和特征融合的单阶段小目标检测算法[J]. 计算机与现代化, 2021, 0(6): 18-23. DOI: 10.3969/j.issn.1006-2475.2021.06.004
作者姓名:王良玮  陈梅  李晖  李焕军  施若  戴震宇
作者单位:贵州省先进计算与医疗信息服务工程实验室(贵州大学计算机科学与技术学院),贵州 贵阳 550025;航天江南集团有限公司,贵州 贵阳 550009;贵州联科卫信科技有限公司,贵州 贵阳 550001
基金项目:国家自然科学基金资助项目(71964009); 贵州省高层次创新型人才项目(黔财教[2018]190)
摘    要:针对经典的单阶段多目标检测算法SSD对小目标物检测效果差的问题,提出DF-SSD算法,其核心技术贡献包括基于反卷积与特征融合的方法和改进后的先验框尺寸计算算法.反卷积与特征融合能够增加浅层特征层的语义信息.改进后的先验框尺寸计算引入了数据集的特点,能有效利用每一个先验框进行训练和预测.改进后的方法DF-SSD与基于SS...

关 键 词:SSD模型  反卷积  特征融合  小目标检测  PASCAL VOC2007  DIOR
收稿时间:2021-07-05

DF-SSD:A One-stage Small Target Detection Algorithm Based on Deconvolution and Feature Fusion
WANG Liang-wei,CHEN Mei,LI Hui,LI Huan-juan,SHI Ruo,DAI Zhen-yu. DF-SSD:A One-stage Small Target Detection Algorithm Based on Deconvolution and Feature Fusion[J]. Computer and Modernization, 2021, 0(6): 18-23. DOI: 10.3969/j.issn.1006-2475.2021.06.004
Authors:WANG Liang-wei  CHEN Mei  LI Hui  LI Huan-juan  SHI Ruo  DAI Zhen-yu
Abstract:Aiming at the problem of the SSD model’s poor detection performance on small targets, the DF-SSD algorithm was proposed, its technical contributions include a one-stage detector method based on deconvolution and feature fusion and an improved default bounding boxes’ size calculation algorithm. Deconvolution and feature fusion can increase the semantic information of shallow feature layers. In DF-SSD algorithm, the improved default bounding boxes’ size calculation introduces the characteristics of the data set, which can effectively use each default bounding box for training and prediction. Compared with the improved R-SSD and RSSD models based on SSD, the DF-SSD method has higher detection accuracy. At the same time, DF-SSD’s detection overhead is only 1/2 of R-SSD and 1/5 of DSSD. The MAP of the DF-SSD on the VOC2007 and DIOR data sets is 1.4 and 3.6 percentage points higher than that of SSD respectively. Meanwhile, DF-SSD’s MAP of small targets of ship, vehicle, windmill, and cat increased 23.2, 12.6, 8 and 4.8 percentage points respectively. The results show DF-SSD effectively improves the detection accuracy of small targets and has a faster detection speed.
Keywords:SSD model   deconvolution   feature fusion   small target detection   PASCAL VOC2007   DIOR  
本文献已被 万方数据 等数据库收录!
点击此处可从《计算机与现代化》浏览原始摘要信息
点击此处可从《计算机与现代化》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号