Image scoring method and system based on deep learning

Abstract

The invention discloses an image scoring method based on deep learning, and the method is characterized in that the method comprises the steps: 1, preparing a picture data set, and carrying out the scoring marking of a picture from different dimensions; 2, training a picture scoring model by using a deep neural network, wherein the picture scoring model is used for evaluating scores of the pictures in more than one designed dimension; 3, carrying out weighted summation on the scores under each dimension to obtain a final score S = sigma(wkScore) of the picture. According to the method, the picture scoring angle of human eyes is referred, the picture scoring model is trained by using deep nerves and the deep neural network, the picture scoring result better meets the human eye judgment standard, the scoring accuracy is greatly improved, the picture screening workload of websites or APPs can be greatly reduced, and the picture click rate of the websites can be increased.

Description

technical field [0001] The present invention relates to the field of computer deep learning, in particular to a method and system for image scoring based on deep learning. Background technique [0002] In all kinds of websites or apps, pictures are the most intuitive form of presentation. Good pictures can quickly attract users' attention, resulting in relatively high clicks and conversions. The daily update volume of pictures is very large, and it is obviously unrealistic to rely on manual scoring. The quality of the picture is very related to the shooting hardware, target environment, shooting skills and other factors. How to automatically, correctly and efficiently rate the pictures is a relatively popular question. [0003] Traditional image scoring methods include mean square error, peak signal-to-noise ratio and other methods, but these methods only reflect the difference in pixels of the image and have nothing to do with the visual observation effect of the human eye...

AI Technical Summary

Problems solved by technology

The amount of updated pictures every day is very large, and it is obviously unrealistic to rely on manual scoring. The quality of pictures is very related to many factors such as shooting hardware, target environment, and shooting skills. How to automatically and efficiently score pictures is a popular question

[0003] Traditional image scoring methods include methods such as mean square error and peak signal-to-noise ratio, but these methods only reflect the difference in pixels of the image and have nothing to do with the visual observation effect of the human eye, so there are certain defects. The observation effect is comprehensively affected by various factors including sharpness, exposure effect, and composition. The existing technology still lacks effective solutions for the selection, analysis, and combination of these factors.

Images