Method and system for identifying uglification pictures of specific characters

Abstract

The invention provides a method and a system for identifying uglification pictures of specific characters. The method comprises the steps of obtaining a picture set containing specific characters, wherein the picture set comprises a plurality of cartoons and a plurality of photos, and each picture in the picture set is marked with a preset label representing whether the picture is uglified or not,taking the cartoons and the photos in the picture set as training data respectively, training a convolutional neural network model, and obtaining a specific character cartoon recognition network anda specific character photo recognition network; judging whether the to-be-identified picture belongs to a cartoon or not through a front-end network; and if so, sending the to-be-identified picture tothe specific character cartoon identification network, obtaining a specific figure uglification picture identification result of the to-be-identified picture; and otherwise, carrying out face comparison on the to-be-identified picture, judging whether the to-be-identified picture comprises the specific character or not, if so, obtaining a specific charater uglification picture identification result of the to-be-identified picture through the specific character picture identification network, and otherwise, obtaining an identification result that the to-be-identified picture does not comprisethe specific character.

Description

technical field [0001] The method belongs to the field of machine learning and computer vision, and in particular relates to machine learning problems oriented to image classification in computer vision. Background technique [0002] With the rapid development of the Internet, the content disseminated on the Internet is no longer limited to text, and the proportion of pictures is increasing. For example, the daily average number of pictures on Weibo is about 9 million. Among such a large number of pictures, there are many pictures that contradict the facts and even slander others, causing adverse effects. At present, the technical research on bad information content is mostly concentrated on text analysis, and there is a lack of key technologies for picture recognition, especially for specific people. A small amount of related work mainly relies on business personnel to combine pictures and surrounding text and other information to carry out visual recognition. This method ...

AI Technical Summary

Problems solved by technology

Among such a large number of pictures, there are many pictures that contradict the facts and even slander others, causing adverse effects

At present, the technical research on bad information content is mostly concentrated on text analysis, and there is a lack of key technologies for picture recognition, especially for specific people. A small amount of related work mainly relies on business personnel to combine pictures and surrounding text and other information to carry out visual recognition. This method The efficiency is low and the discovery ability is limited, so it is necessary to use the method of machine learning to realize the automatic recognition of specific person pictures

Images