General Image Aesthetics Evaluation Method, System and Equipment Based on Attention Mechanism

Abstract

The present invention relates to the technical fields of image recognition and machine learning, in particular to a method, system and equipment for evaluating general image aesthetics based on an attention mechanism, with the purpose of improving the accuracy of evaluation results. The evaluation method of the present invention includes: scaling the image to be evaluated, so that the length of the shortest side of the scaled image is equal to the preset first length; randomly cutting a preset number of square image blocks on the scaled image, making The side length of the square image block is equal to the preset second length; each square image block is input into the trained convolutional neural network model, and a corresponding two-dimensional aesthetic level confidence vector is output; the preset number of The mean value of the two-dimensional aesthetic level confidence vector; perform aesthetic evaluation on the image to be evaluated according to the mean value. The accuracy rate of the invention is obviously higher than that of the existing technical solutions; no additional image information is needed; the evaluation process consumes less time and the model occupies less space.

Description

technical field [0001] The invention relates to the technical fields of image recognition and machine learning, in particular to a method, system and equipment for general image aesthetics evaluation based on attention mechanism. Background technique [0002] General image aesthetic evaluation aims to use computer systems to intelligently judge the aesthetic level of input images, and requires that the judgment given by the system has a high consistency with the judgment made by human experts with good aesthetic taste. General image aesthetic evaluation is one of the foundations of many technologies such as image recommendation and image post-processing, and it is also an interdisciplinary subject (including cognitive psychology, computer vision, machine learning, etc.). The level of is an important issue of attention and investment. [0003] The current mainstream general image aesthetic evaluation methods rely on the use of additional image information (such as object cat...

AI Technical Summary

Problems solved by technology

This type of design method will bring a large amount of training, and it is not an end-to-end (End-to-End) training mode, and cannot effectively complete the training task from the data level

[0008] Both of the above two solutions require a lot of manpower to label the additional information of the image, and the types of additional information depend on expert design, which is not only time-consuming and labor-intensive, but also difficult to maintain and expand

Images