基于Grad-CAM与KL损失的SSD目标检测算法

doi:10.3969/j.issn.0372-2112.2020.12.017

电子学报 ›› 2020, Vol. 48 ›› Issue (12): 2409-2416.DOI: 10.3969/j.issn.0372-2112.2020.12.017

基于Grad-CAM与KL损失的SSD目标检测算法

侯庆山, 邢进生

山西师范大学数学与计算机科学学院, 山西临汾 041004

收稿日期:2019-11-14 修回日期:2020-04-14 出版日期:2020-12-25
- 通讯作者:
- 邢进生
- 作者简介:
- 侯庆山 男,1995年出生于山西阳泉,现为山西师范大学数学与计算机科学学院在读硕士研究生.主要研究方向:计算机视觉.E-mail:18435202842@163.com
- 基金资助:
- 山西省软科学基金资助项目 (No.2011041033-03）

SSD Object Detection Algorithm Based on KL Loss and Grad-CAM

HOU Qing-shan, XING Jin-sheng

School of Mathematics and Computer Science, Shanxi Normal University, Linfen, Shanxi 041004, China

Received:2019-11-14 Revised:2020-04-14 Online:2020-12-25 Published:2020-12-25

摘要/Abstract

摘要： 鉴于Single Shot Multibox Detector（SSD）算法对中小目标检测时会出现漏检甚至错检的情况，提出一种改进的SSD目标检测算法，以提高中小目标检测的准确性.运用Gradient-weighted Class Activation Mapping（Grad-CAM）技术对检测过程中的细节作可视化处理，并以类激活图的形式呈现各检测层细节，分析各检测层的类激活图发现SSD算法中待检测目标的错检以及中小目标的漏检现象与回归损失函数相关.据此，采用Kullback-Leibler（KL）边框回归损失策略，利用Non Maximum Suppression（NMS）算法输出最终预测框.实验结果表明，改进算法相较于已有检测算法具有更高的准确率以及稳定性.

关键词: 目标检测, 可视化, 类激活图, Grad-CAM, SSD, KL损失

Abstract: Considering that the single shot multibox detector (SSD) algorithm will be missed or even false when it is used to detect the small and medium-sized objects,an improved SSD object detection algorithm is proposed to improve the accuracy of small and medium-sized objects detection.The details in the detection process are visualized with gradient-weighted class activation mapping (Grad-CAM) technology,and the details of each detection layer are shown in the form of class activation maps.Then it is noted that the phenomenon of the false or missed detection of the objects to be detected on small and medium-sized objects in the SSD algorithm is related to the regression loss function.Accordingly,Kullback-Leibler (KL) border regression loss strategy is adopted and non maximum suppression (NMS) algorithm is used to output the final prediction boxes.Experimental results show that compared with the existing detection algorithms,the improved algorithm in this paper has higher accuracy and stability.

Key words: object detection, visualization, class activation maps, Grad-CAM, SSD, KL loss

中图分类号:

TP183

侯庆山, 邢进生. 基于Grad-CAM与KL损失的SSD目标检测算法[J]. 电子学报, 2020, 48(12): 2409-2416.

HOU Qing-shan, XING Jin-sheng. SSD Object Detection Algorithm Based on KL Loss and Grad-CAM[J]. Acta Electronica Sinica, 2020, 48(12): 2409-2416.

参考文献

[1] Gomaa Ahmed,Abdelwahab Moataz M,et al.Robust vehicle detection and counting algorithm employing a convolution neural network and optical flow[J].Sensors (Basel,Switzerland),2019,19(20):4588-4600.
[2] Ju J,Xing J.Moving object detection based on smoothing three frame difference method fused with RPCA[J].Multimed Tools Appl,2019,78(21):29937-29951.
[3] You Zhou,Ting Rui,Yiran Li,Xuegang Zuo.A UAV patrol system using panoramic stitching and object detection[J].Computers and Electrical Engineering,2019,80(17):106473-106481.
[4] Shen Zhao,Xi Wu,Bo Chen,Shuo Li.Automatic spondylolisthesis grading from MRIs across modalities using faster adversarial recognition network[J].Medical Image Analysis,2019,58(12):101533-101572.
[5] Redmon J,Divvala S,Girshick R,Farhadi A.You only look once:Unified,real-time object detection[A].2016 IEEE Conference on Computer Vision and Pattern Recognition[C].Las Vegas:IEEE,2016.779-788.
[6] Liu W,Anguelov D,Erhan D,et al.SSD:Single shot multibox detector[A].Computer Vision-ECCV 2016-14th European Conference[C].Amsterdam:IEEE,2016.21-37.
[7] Girshick R,Donahue J,Darrell T,Malik J.Rich feature hierarchies for accurate object detection and semantic segmentation[A].2014 IEEE Conference on Computer Vision and Pattern Recognition[C].Columbus:IEEE,2014.580-587.
[8] 温捷文,战荫伟,李楚宏.一种加强SSD小目标检测能力的Atrous滤波器设计[J].计算机应用研究,2019,36(3):861-865. Wen J W,Zhan Y W,Li C H,et al.Design of Atrous filter to strengthen small object detection capability of SSD[J].The Pascal Application Research of Computers,2019,36(3):861-865.
[9] Wu Tianshu,Zhang Zhijia,Liu Yunpeng,et al.A lightweight small object detection algorithm based on improved SSD[J].Infrared and Laser Engineering,2018,47(7):703005.
[10] 邢浩强,杜志岐,苏波.基于改进SSD行人检测方法[J].计算机工程,2018,44(11):228-233. Xing H Q,Du Z Q,Su B.Pedestrian detection method based on modified SSD[J].Computer Engineering,2018,44(11):228-233.(in Chinese)
[11] Tang C,Ling Y S,Zheng K D,et al.Object detection method of multi-view SSD based on deep learning[J].Infrared and Laser Engineering,2018,47(1):302-310.
[12] Chen H J,Wang Q Q,Yang G W,et al.SSD object detection algorithm with multi-scale convolution feature fusion[J].Journal of Frontiers of Computer Science and Technology,2019,13(6):1049-1061.
[13] Fu C Y,Liu W,Ranga A,et al.DSSD:Deconvolutional Single Shot Detector[EB/OL].https://arxiv.org/abs/1701.06659,2017-01-23.
[14] Jeong J,Park H,Kwak N.Enhancement of SSD by Concatenating Feature Maps for Object Detection[EB/OL].https://arxiv.org/abs/1705.09587,2017-05-26.
[15] Li Z X,Zhou F Q.FSSD:Feature Fusion Single Shot Multibox Detector[EB/OL].https://arxiv.org/abs/1712.00960,2018-05-17.
[16] Zeiler M D,Fergus R.Visualizing and understanding convolutional neural networks[A].Computer Vision-ECCV 2014-13th European Conference[C].Zurich:IEEE,2014.818-833.
[17] K Simonyan,A Zisserman.Very Deep Convolutional Networks for Large-scale Image Recognition[EB/OL].https://arxiv.org/abs/1409.1556,2015-04-10.
[18] Lin T Y,Dollár,Piotr,Girshick R,et al.Feature pyramid networks for object detection[A].2017 IEEE Conference on Computer Vision and Pattern Recognition[C].Honolulu:IEEE,2017.2117-2125.
[19] Y He,C Zhu,J Wang,M Savvides,X Zhang.Bounding box regression with uncertainty for accurate object detection[A].2019 IEEE Conference on Computer Vision and Pattern Recognition[C].Long Beach:IEEE,2019.2888-2897.
[20] Ren S,He K,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2015,39(6):1137-1149.
[21] K He,G Gkioxari,P Dollar,R Girshick.Mask R-CNN[EB/OL].https://arxiv.org/abs/1703.06870,2018-01-24.
[22] Ramprasaath R,Abhishek D,Ramakrishna V,Michael C,Devi P,Dhruv B.Grad-CAM:Why did you say that? visual explanations from deep networks via gradient-based localization[A].2017 IEEE Conference on Computer Vision and Pattern Recognition[C].Honolulu:IEEE,2017.618-626.
[23] Farahnak-Ghazani F,Baghshah M S.Multi-label classification with feature-aware implicit encoding and generalized cross-entropy loss[A].24th Iranian Conference on Electrical Engineering[C].Persia:IEEE,2016.7585772.
[24] J Redmon,A Farhadi.YOLO9000:better,faster,stronger[A].2017 IEEE Conference on Computer Vision and Pattern Recognition[C].Honolulu:IEEE,2017.7263-7271.
[25] Girshick R.Fast R-CNN[A].2015 IEEE International Conference on Computer Vision[C].Santiago:IEEE,2015.1440-1448.

基于Grad-CAM与KL损失的SSD目标检测算法

SSD Object Detection Algorithm Based on KL Loss and Grad-CAM

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	琚长瑞, 秦晓燕, 袁广林, 李豪, 朱虹. 尺度敏感损失与特征融合的快速小目标检测方法[J]. 电子学报, 2022, 50(9): 2119-2126.
[2]	李超, 黄新宇, 王凯. 基于特征融合和自学习锚框的高分辨率图像小目标检测算法[J]. 电子学报, 2022, 50(7): 1684-1695.
[3]	包中华, 卢建斌, 田永华, 田树森. 密集杂波背景下雷达微弱海面目标的修正Hough变换TBD检测新方法[J]. 电子学报, 2022, 50(7): 1735-1743.
[4]	肖进胜, 张舒豪, 陈云华, 王元方, 杨力衡. 双向特征融合与特征选择的遥感影像目标检测[J]. 电子学报, 2022, 50(2): 267-272.
[5]	李斌, 齐延荣, 周清雷. 基于Winograd算法的目标检测加速器设计与优化[J]. 电子学报, 2022, 50(10): 2387-2397.
[6]	罗会兰, 袁璞, 童康. 基于深度学习的显著性目标检测方法综述[J]. 电子学报, 2021, 49(7): 1417-1427.
[7]	程旭, 宋晨, 史金钢, 周琳, 张毅锋, 郑钰辉. 基于深度学习的通用目标检测研究综述[J]. 电子学报, 2021, 49(7): 1428-1438.
[8]	周勇, 陈思霖, 赵佳琦, 张迪, 王瀚正. 基于弱语义注意力的遥感图像可解释目标检测[J]. 电子学报, 2021, 49(4): 679-689.
[9]	王洪雁, 张海坤, 罗宇华, 汪祖民. 复杂动态背景下基于群稀疏的运动目标检测方法[J]. 电子学报, 2021, 49(12): 2330-2338.
[10]	罗大鹏, 杜国庆, 曾志鹏, 魏龙生, 高常鑫, 陈应, 肖菲, 罗琛. 基于少量样本学习的多目标检测跟踪方法[J]. 电子学报, 2021, 49(1): 183-191.
[11]	李维刚, 叶欣, 赵云涛, 王文波. 基于改进YOLOv3算法的带钢表面缺陷检测[J]. 电子学报, 2020, 48(7): 1284-1292.
[12]	罗会兰, 陈鸿坤. 基于深度学习的目标检测研究综述[J]. 电子学报, 2020, 48(6): 1230-1239.
[13]	刘颖, 刘红燕, 范九伦, 公衍超, 李莹华, 王富平, 卢津. 基于深度学习的小目标检测研究与应用综述[J]. 电子学报, 2020, 48(3): 590-601.
[14]	孟祥伟. 量化秩非参数CFAR检测器在杂波边缘中的性能分析[J]. 电子学报, 2020, 48(2): 384-389.
[15]	权宇, 李志欣, 张灿龙, 马慧芳. 融合深度扩张网络和轻量化网络的目标检测模型[J]. 电子学报, 2020, 48(2): 390-397.