Online Progressive Instance-Balanced Sampling for Weakly Supervised Vibration Damper Detection

doi:10.1109/TIM.2023.3273655

CORC > 自动化研究所 > 中国科学院自动化研究所 > 精密感知与控制研究中心 > 精密感知与控制

	Online Progressive Instance-Balanced Sampling for Weakly Supervised Vibration Damper Detection
	Chen, Minghao 1,2; Tian, Yunong 1,2; Li, Zhishuo 1,2; Li, En 3; Liang, Zize 1,2
刊名	IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
	2023
卷号	72 页码:14
关键词	Shock absorbers Vibrations Object detection Proposals Training Sampling methods Convolutional neural networks Instance balance multiple instance learning (MIL) progressive sampling vibration damper detection weakly supervised object detection (WSOD)
ISSN号	0018-9456
DOI	10.1109/TIM.2023.3273655
通讯作者	Li, En(en.li@ia.ac.cn)
英文摘要	The detection of vibration dampers is important for power systems. Deep learning-based damper detection needs massive annotations, which are labor-intensive and time-consuming. Therefore, weakly supervised object detection (WSOD) is considered. Part domination is the main problem in weakly supervised vibration damper detection. Current WSOD methods neglect that overwhelming negative instances exist in each image during the training phase, which would mislead the training and make detection results stuck in the most discriminative parts of objects. To tackle this problem, an online progressive instance-balanced sampling (OPIS) algorithm based on hard sampling and soft sampling is proposed in this article. The algorithm includes two modules: a progressive instance balance (PIB) module and a progressive instance reweighting (PIR) module. The PIB module, combining random sampling and intersection over union (IoU)-balanced sampling, progressively mines hard negative instances while balancing positive instances and negative instances. The PIR module further utilizes classifier scores and IoUs of adjacent refinements to reweight the weights of positive instances to make the network focus on positive instances. Extensive experimental results on the vibration damper, pattern analysis, statistical modelling and computational learning visual object classes (PASCAL VOC) 2007, and PASCAL VOC 2012 datasets demonstrate that the proposed method can significantly improve the baseline, which is also comparable to many existing methods. In addition, compared with the baseline, the proposed method requires no extra network parameters, and the supplementary training overheads are small.
资助项目	National Natural Science Foundation of China[62273344] ; National Key Research and Development Program of China[2018YFB1307400]
WOS研究方向	Engineering ; Instruments & Instrumentation
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000994621200002
资助机构	National Natural Science Foundation of China ; National Key Research and Development Program of China
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53561]
专题	精密感知与控制研究中心_精密感知与控制
通讯作者	Li, En
作者单位	1.Univ Chinese Academyof Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China 2.Inst Automation, Chinese Acad Sci, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China 3.Chinese Acad Sci, Engn Lab Ind Vis & Intelligent Equipment Technol, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Chen, Minghao,Tian, Yunong,Li, Zhishuo,et al. Online Progressive Instance-Balanced Sampling for Weakly Supervised Vibration Damper Detection[J]. IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT,2023,72:14.
APA	Chen, Minghao,Tian, Yunong,Li, Zhishuo,Li, En,&Liang, Zize.(2023).Online Progressive Instance-Balanced Sampling for Weakly Supervised Vibration Damper Detection.IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT,72,14.
MLA	Chen, Minghao,et al."Online Progressive Instance-Balanced Sampling for Weakly Supervised Vibration Damper Detection".IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 72(2023):14.