Content-Based Image Retrieval using Deep Convolutional Neural Networks

Image retrieval is an important issue of machine vision and image processing. Many researches have been done in image retrieval. In 70’s, Text-Based image retrieval had been created before Content-Based image retrieval have been introduced since 90’s cause of large amount of data stored and inefficient previous methods. On this way, researcher reached better conclusion by extracting features from pictures. Semantic gap between these features and human concept, and burst increase in amount of images which were saved, caused researchers to think about new algorithms. Excellent successes on deep learning algorithms encourage us to implant a new method for image retrieval based on deep learning. In this paper, after reviewing deep convolutional neural networks as a kind of deep learning methods, we introduce a new retrieval system based on deep convolutional neural networks and by testing it on three famous databases, ALOI, Corel and MPEG7, computing P(0.5), P(1) and ANMRR and comparing them with other methods which have been used since recent years, we show the superior accuracy of this method in comparison to the other methods.