Large-scale image retrieval method and system based on deep convolutional neural network

Abstract

The invention discloses a large-scale image retrieval method and system based on a deep convolutional neural network (DHN), large-scale image retrieval is carried out by constructing the deep convolutional neural network (DHN), and the DHN comprises ResNet50-based feature extraction, channel and space attention (CSA)-based feature refinement, a classification layer, a hash layer and a weight layer. The DHN realizes pixel saliency attention from bottom to top through the CSA, and realizes semantic constraint from top to bottom through classification label supervision; the DHN adopts an adaptive weighted learning algorithm to generate a weight for each bit of Hash code, and then directly generates a short Hash code from a long Hash code according to the importance of the bit represented by the weight. The method provided by the invention has higher Hash code generation precision and speed, thereby being suitable for large-scale image retrieval tasks.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and relates to a large-scale image retrieval method and system, in particular to a large-scale image retrieval method and system based on mixed attention and adaptive weights. Background technique [0002] The task of large-scale image retrieval is to quickly find a certain number of images with similar content from a million-level image database given a query image through image features. Traditional tree-based retrieval methods usually use high-time-complexity similarity measures such as Euclidean distance to calculate the distance between features, and show good performance when dealing with low-dimensional data. However, when the amount of data reaches millions or hundreds of millions, and the feature dimension grows by a large margin, dimensional disaster is prone to occur, and the time performance is significantly degraded. In order to be able to significantly improve the s...

AI Technical Summary

Problems solved by technology

[0008] Existing deep hashing methods have achieved high retrieval accuracy, but most methods only extract the low-level features of the picture, and the extracted features are easily disturbed by irrelevant objects in the picture, resulting in hash codes generated by similar data points. are not similar, so the retrieval accuracy and robustness need to be improved

Images