Visual emotion recognition method

Abstract

The invention discloses a visual emotion recognition method, and the method comprises the steps: extracting the image feature representation of context awareness from image data, wherein the image feature representation comprises the global feature information of a whole image and the local feature information of an area where an interested person is located; performing joint modeling on the emotion state label of the interested person contained in the image data and the extracted global feature information of the whole picture by utilizing an attention mechanism and adopting a GRU unit; and combining a joint modeling result with the local feature information of the region where the interested person is located to obtain a preliminarily predicted emotional state label, constructing an output filter, obtaining probability distribution of the preliminarily predicted emotional state label on all categories, and finally calculating a predicted emotional state label category of each time step. According to the method, global and local information is combined, so that more comprehensive feature representation is obtained; in addition, the method provided by the invention carries out joint training on the labels and the image features, so that the problems of relevance and cooccurrence between the labels are solved.

Description

technical field [0001] The invention relates to the fields of deep learning, visual emotion recognition and multi-label classification, in particular to a visual emotion recognition method. Background technique [0002] Visual Emotion Recognition (Visual Emotion Recognition) is a particularly important component and research field in the field of computer vision. Its task is to classify the emotional state of the characters in a given picture, which has a wide range of application scenarios in many fields, such as Intelligent Internet of Things (IoT), online education, video sentiment analysis, political election prediction, etc. [0003] At present, the methods of visual emotion classification mainly include the following methods: [0004] 1) A method based on facial expression recognition. [0005] The method based on facial expression recognition mainly uses traditional feature engineering methods, or combines deep learning with different neural network structures to ex...

AI Technical Summary

Problems solved by technology

In addition, the method of the present invention jointly models tags and image features, so as to solve the problem of correlation and co-occurrence between tags

Images