Image recognition method based on double attention

Abstract

The invention discloses an image recognition method based on double attention. The method comprises the following steps of creating a double attention VLAD network model comprising a convolution layer, a spatial attention VLAD layer, a channel attention VLAD layer, a mixed error function and a full connection layer; inputting a to-be-identified image into a dual-attention VLAD network model to enable a convolution layer in the dual-attention VLAD network model to output a first feature map; inputting the first feature map into a spatial attention VLAD layer and a channel attention VLAD layer to obtain a first VLAD feature vector and a second VLAD feature vector respectively; carrying out feature fusion on the first VLAD feature vector and the second VLAD feature vector and then inputting the fused vectors into a full connection layer; and outputting a recognition result of the to-be-recognized image, thereby solving a technical problem that the recognition precision of a network modelis not high because the network model introduced with VLAD in the conventional neural network image recognition is liable to be interfered by low-quality image features.

Description

technical field [0001] The present application relates to the technical field of image recognition, in particular to an image recognition method based on double attention. Background technique [0002] In the field of image recognition research, in order to obtain a better description of image information and achieve better recognition results, it is mainly carried out from two aspects: traditional image processing methods and new neural network image recognition technology. [0003] Traditional artificial design feature extraction methods are highly targeted for features. Therefore, many feature extraction and description methods have better recognition results. Traditional methods use more manual feature extraction, such as Sift, BoW, FV and VLAD (Vector of Locally Aggregated Descriptors, local feature aggregation descriptors), etc., have a more detailed description of the local information of the image, and the amount of calculation is less. However, the disadvantage of t...

AI Technical Summary

Problems solved by technology

[0005] This application provides an image recognition method based on double attention, which is used to solve the problem that the network model introduced by VLAD in the existing neural network image recognition is easily disturbed by low-quality image features, so that the recognition accuracy of the network model is not high. question

Images