Image similarity determination method and device, equipment and storage medium

Abstract

The invention discloses an image similarity determination method. The method comprises the following steps: obtaining a pair of target images to be determined in similarity; the pair of target imagescomprises a first image and a second image; the pair of target images comprises a first image and a second image. Extracting at least two image features of a target image pair; Each image feature of the target image pair is input into a depth convolution network model trained in advance based on the corresponding image feature, and the similarity determination results of each depth convolution network model are obtained. The similarity between the first image and the second image in the target image pair is determined based on the weighting operation of the similarity determination result of each depth convolution network model. The technical proposal provided by the embodiment of the invention effectively improves the determination accuracy of the image similarity, and improves the performance of the image recognition and the target tracking task based on the image similarity. The invention also discloses an image similarity determination device, equipment and a storage medium, whichhave corresponding technical effects.

Description

technical field [0001] The present invention relates to the field of computer application technology, in particular to a method, device, equipment and storage medium for determining image similarity. Background technique [0002] In application scenarios such as image recognition and object tracking, it is often necessary to determine the similarity between two images. [0003] In the prior art, Siamese networks can be used to determine image similarity. The Siamese network is a special deep convolutional network structure. The network has two data input ports. The convolutional network structure can be used to convolve the image data of the two ports at the same time, and finally extract the feature vector of the input image. Then use the eigenvectors of the two images and a specific similarity measurement function to further determine the similarity. [0004] This method uses a single network model to judge the similarity of two images, and the judgment result is highly ...

AI Technical Summary

Problems solved by technology

[0004] This method uses a single network model to judge the similarity of two images, and the judgment result is highly dependent on the reliability of the network model training. In this network model, a single image feature is used as input, and there is a certain one-sidedness in the image representation. As a result, the accuracy of determining the image similarity is not high

Images