The Application of Improved YOLO V3 in Multi-Scale Target Detection

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 光电信息技术研究室

	The Application of Improved YOLO V3 in Multi-Scale Target Detection
	Luo HB(罗海波)1,2,4,5 ; Hui B(惠斌)1,2,4,5 ; Chang Z(常铮)1,2,4,5 ; Ju MR(鞠默然)1,2,3,4,5 ; Wang ZB(王仲博)1,2,3,4,5
刊名	APPLIED SCIENCES-BASEL
	2019
卷号	9 期号:18 页码:1-14
关键词	target detection YOLO V3 anchor box computer vision deep learning
ISSN号	2076-3417
产权排序	1
英文摘要	Target detection is one of the most important research directions in computer vision. Recently, a variety of target detection algorithms have been proposed. Since the targets have varying sizes in a scene, it is essential to be able to detect the targets at different scales. To improve the detection performance of targets with different sizes, a multi-scale target detection algorithm was proposed involving improved YOLO (You Only Look Once) V3. The main contributions of our work include: (1) a mathematical derivation method based on Intersection over Union (IOU) was proposed to select the number and the aspect ratio dimensions of the candidate anchor boxes for each scale of the improved YOLO V3; (2) To further improve the detection performance of the network, the detection scales of YOLO V3 have been extended from 3 to 4 and the feature fusion target detection layer downsampled by 4x is established to detect the small targets; (3) To avoid gradient fading and enhance the reuse of the features, the six convolutional layers in front of the output detection layer are transformed into two residual units. The experimental results upon PASCAL VOC dataset and KITTI dataset show that the proposed method has obtained better performance than other state-of-the-art target detection algorithms.
语种	英语
WOS记录号	WOS:000489115200135
内容类型	期刊论文
源URL	[http://ir.sia.cn/handle/173321/25771]
专题	沈阳自动化研究所_光电信息技术研究室
作者单位	1.The Key Lab of Image Understanding and Computer Vision, Shenyang 110016, China 2.Key Laboratory of Opto-Electronic Information Processing, Chinese Academy of Science, Shenyang 110016, China 3.University of Chinese Academy of Sciences, Beijing 100049, China 4.Institutes for Robotics and IntelligentManufacturing, Chinese Academy of Sciences, Shenyang 110169, China 5.Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China
推荐引用方式 GB/T 7714	Luo HB,Hui B,Chang Z,et al. The Application of Improved YOLO V3 in Multi-Scale Target Detection[J]. APPLIED SCIENCES-BASEL,2019,9(18):1-14.
APA	Luo HB,Hui B,Chang Z,Ju MR,&Wang ZB.(2019).The Application of Improved YOLO V3 in Multi-Scale Target Detection.APPLIED SCIENCES-BASEL,9(18),1-14.
MLA	Luo HB,et al."The Application of Improved YOLO V3 in Multi-Scale Target Detection".APPLIED SCIENCES-BASEL 9.18(2019):1-14.