Image retrieval method based on metric learning for convolutional neural network

2017 2nd International Seminar on Advances in Materials Science and Engineering

Wang, Jieyuan^1 ; Qian, Ying^1 ; Ye, Qingqing^1 ; Wang, Biao^1

School of Software Engineering, Chongqing University of Posts and Telecommunications, Chongqing

400065, China^1

关键词: Approximate nearest neighbors (ANN);    Content-Based Image Retrieval;    Convolutional networks;    Convolutional neural network;    High dimension problems;    Retrieval performance;    Similarity measure;    Visual feature extraction;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/231/1/012002/pdf
DOI  :  10.1088/1757-899X/231/1/012002

来源: IOP

【 摘 要 】

At present, the research of content-based image retrieval (CBIR) focuses on learning effective feature for the representations of origin images and similarity measures. The retrieval accuracy and efficiency are crucial to a CBIR. With the rise of deep learning, convolutional network is applied in the domain of image retrieval and achieved remarkable results, but the image visual feature extraction of convolutional neural network exist high dimension problems, this problem makes the image retrieval and speed ineffective. This paper uses the metric learning for the image visual features extracted from the convolutional neural network, decreased the feature redundancy, improved the retrieval performance. The work in this paper is also a necessary part for further implementation of feature hashing to the approximate-nearest-neighbor (ANN) retrieval method.