A Two-Stage Triplet Network Training Framework for Image Retrieval

doi:10.1109/TMM.2020.2974326

	A Two-Stage Triplet Network Training Framework for Image Retrieval
	Min, Weiqing 1,2; Mei, Shuhuan 3; Li, Zhuo 1,2; Jiang, Shuqiang 1,2
刊名	IEEE TRANSACTIONS ON MULTIMEDIA
	2020-12-01
卷号	22 期号:12 页码:3128-3138
关键词	Image retrieval Feature extraction Convolution Training Image representation Task analysis Measurement
ISSN号	1520-9210
DOI	10.1109/TMM.2020.2974326
英文摘要	In this paper, we propose a novel framework for instance-level image retrieval. Recent methods focus on fine-tuning the Convolutional Neural Network (CNN) via a Siamese architecture to improve off-the-shelf CNN features. They generally use the ranking loss to train such networks, and do not take full use of supervised information for better network training, especially with more complex neural architectures. To solve this, we propose a two-stage triplet network training framework, which mainly consists of two stages. First, we propose a Double-Loss Regularized Triplet Network (DLRTN), which extends basic triplet network by attaching the classification sub-network, and is trained via simultaneously optimizing two different types of loss functions. Double-loss functions of DLRTN aim at specific retrieval task and can jointly boost the discriminative capability of DLRTN from different aspects via supervised learning. Second, considering feature maps of the last convolution layer extracted from DLRTN and regions detected from the region proposal network as the input, we then introduce the Regional Generalized-Mean Pooling (RGMP) layer for the triplet network, and re-train this network to learn pooling parameters. Through RGMP, we pool feature maps for each region and aggregate features of different regions from each image to Regional Generalized Activations of Convolutions (R-GAC) as final image representation. R-GAC is capable of generalizing existing Regional Maximum Activations of Convolutions (R-MAC) and is thus more robust to scale and translation. We conduct the experiment on six image retrieval datasets including standard benchmarks and recently introduced INSTRE dataset. Extensive experimental results demonstrate the effectiveness of the proposed framework.
资助项目	National Natural Science Foundation of China[61532018] ; National Natural Science Foundation of China[61972378] ; Beijing Natural Science Foundation[L182054] ; National Program for Special Support of Eminent Professionals ; National Program for Support of Top-notch Young Professionals
WOS研究方向	Computer Science ; Telecommunications
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000591817700009
内容类型	期刊论文
源URL	[http://119.78.100.204/handle/2XEOYT63/16368]
专题	中国科学院计算技术研究所
通讯作者	Jiang, Shuqiang
作者单位	1.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 2.Chinese Acad Sci, Inst Comp Technol, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China 3.Nanjing New Generat Artificial Intelligence Res I, Nanjing 210046, Peoples R China
推荐引用方式 GB/T 7714	Min, Weiqing,Mei, Shuhuan,Li, Zhuo,et al. A Two-Stage Triplet Network Training Framework for Image Retrieval[J]. IEEE TRANSACTIONS ON MULTIMEDIA,2020,22(12):3128-3138.
APA	Min, Weiqing,Mei, Shuhuan,Li, Zhuo,&Jiang, Shuqiang.(2020).A Two-Stage Triplet Network Training Framework for Image Retrieval.IEEE TRANSACTIONS ON MULTIMEDIA,22(12),3128-3138.
MLA	Min, Weiqing,et al."A Two-Stage Triplet Network Training Framework for Image Retrieval".IEEE TRANSACTIONS ON MULTIMEDIA 22.12(2020):3128-3138.