Semantic image retrieval method based on attention mechanism

Abstract

The invention discloses a semantic image retrieval method based on an attention mechanism, and mainly solves the problem that a semantic gap affects retrieval accuracy in a picture retrieval process.The method comprises the following steps: 1) constructing a CNN-RNN network model containing an attention mechanism and training the CNN-RNN network model; 2) extracting text features of pictures in an image library by using the trained network model; 3) extracting semantic feature vectors of the text features by using a text vector doc2vec model and storing the semantic feature vectors; 4) extracting text features of the query picture by using the trained network model, and extracting semantic feature vectors corresponding to the text features; and 5) calculating and comparing the feature vector of the query picture with the feature vector in the image library by using a cosine method, and outputting a result. According to the method, the influence caused by a semantic gap can be effectively reduced, so that the system can perform similarity retrieval on semantic information shown by the pictures, and the method can be used for quick retrieval of mass data in the Internet and search of mobile phone pictures in daily life.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to an image-based pattern recognition technology, specifically a semantic image retrieval method based on an attention mechanism. During the image retrieval process, for a query image (query image), an image similar to the query image in the image library is searched and output. Background technique [0002] Image retrieval refers to given an image containing specific content, and then finds images containing similar content in the image database, but because different images are quite different under the influence of factors such as shooting angle, occlusion, and illumination, how to solve the above-mentioned impossible problems? Finding the desired image quickly under the influence of control factors is a challenging problem. In today's Internet era, a huge amount of images are uploaded to the server every moment on the Internet, especially with the rise of social ...

AI Technical Summary

Problems solved by technology

Therefore, many improved algorithms appeared on the basis of CNN+hash coding. However, this method also has shortcomings, that is, the problem of "semantic gap" in image retrieval has not been completely solved, that is, it cannot realize the image semantics. angle to retrieve similar images

The disadvantage of this method is that the retrieval system is too complex and there are too many types of features, which will greatly affect the speed of retrieval, and cannot effectively overcome or reduce the "semantic gap" problem in the retrieval process

Images