Image retrieval method and system based on Siamese network

Abstract

The invention provides an image retrieval method and system based on a Siamese network. The method comprises the first step of inputting an image to be retrieved to a trained single-branch Siamese network so that deep network features of the image to be retrieved can be extracted through the single-branch Siamese network; the second step of calculating the similarity between the deep network features of the image to be retrieved and the deep network features of each template image in a database, wherein the similarity represents the similarity between the image to be retrieved and each template image; the third step of adopting a template image which has the largest similarity with the image to be retrieved as a retrieval result image corresponding to the image to be retrieved. According to the image retrieval method and system based on the Siamese network, by constructing the Siamese network and training the Siamese network, and extracting the deep network features of the image through the trained the Siamese network, the extracted features have very strong discriminability, so that the Siamese network has very high accuracy when being used in the field of image retrieval.

Description

technical field [0001] The present invention relates to the technical field of image retrieval, and more specifically, relates to a Siamese network-based image retrieval method and system. Background technique [0002] With the increasing popularity of electronic devices such as mobile phones and cameras and the rapid development of Internet and computer technology, picture and video data has exploded and reached a massive scale. How to effectively utilize rich multimedia resources is a challenge and an urgently needed key technology in the era of big data. Content-based image search technology is one of the key technologies for deep utilization and management of massive image data, and it is also a frontier topic in the field of computer vision and multimedia. Not only that, image search technology also has important applications in real life. At present, we are stepping into the security society. Cameras have spread all over the place in life. Image search can find speci...

AI Technical Summary

Problems solved by technology

However, image retrieval is different from image classification. Image retrieval requires precise matching. Since there are many types of objects, and there are many objects of each type, there are many template images in the process of open image matching retrieval. Therefore, for each template There are relatively few features extracted from images, and there are great challenges in the accuracy of image retrieval

Images