Multi-dimensional aesthetic quality evaluation method and device for mobile game images and medium

Abstract

The invention provides a multi-dimensional aesthetic quality evaluation method and device for mobile game images and a medium, and the method comprises the steps: carrying out the aesthetic feature extraction of the mobile game images through employing a feature extraction network based on deep learning, and carrying out the fusion of the extracted aesthetic features; connecting a full connection layer to the feature extraction network, and mapping the fused aesthetic features to different aesthetic dimensions through the full connection layer to obtain corresponding aesthetic scores, wherein the feature extraction network and the full connection layer form a quality evaluation model; training the quality evaluation model through multi-task learning, so that the quality evaluation model can predict multi-dimensional aesthetic features of the mobile game image; and inputting a to-be-evaluated mobile game image into the trained quality evaluation model to obtain a multi-dimensional aesthetic quality score. According to the method, the aesthetic quality of the mobile game image can be effectively evaluated.

Description

technical field [0001] The present invention relates to the technical field of multimedia quality evaluation, in particular to a multi-dimensional aesthetic quality evaluation method, device and medium for mobile game images based on deep learning. Background technique [0002] In recent years, the global game industry has developed rapidly. With the expansion of mobile Internet applications, mobile games have become the largest and fastest growing segment of the global game market. The official launch of 5G commercial use will promote the rapid development of cloud games, e-sports, VR and other industries with its high-speed and convenient network foundation. In order to keep up with the development trend of the game industry, game developers need to continuously improve the quality of mobile games to gain an advantage in the competition. The aesthetic evaluation of mobile game screens is of great help to the optimization of the quality of game content, the innovation of ...

AI Technical Summary

Problems solved by technology

[0005] Although these methods have achieved very good results, there is still a lack of multi-dimensional aesthetic quality evaluation methods for mobile game images. The previous methods lack good generalization performance in the evaluation of mobile game images. The main reason is that the previous methods The method is mainly aimed at the aesthetic quality of natural scenes, and the content of mobile game images is mostly obtained by virtual rendering, so it is difficult to directly apply the previous methods to mobile game images; in addition, most methods only use subjective evaluation. The MOS score guides the design of objective methods, and does not make full use of the results of subjective evaluation. In fact, the results of subjective evaluation contain richer information;

[0006] All in all, the current objective methods for multi-dimensional aesthetic quality evaluation of mobile phone images are very lacking.

Images