Image sentiment analysis method based on multi-task learning mode

Abstract

The invention discloses an image sentiment analysis method based on a multi-task learning mode. The method comprises the following steps: constructing an image sentiment attribute detector and an image sentiment label classifier; using a gradient descent method to train initialization parameters of the image emotion attribute detector; testing the prediction precision of the emotion attributes of the image and judging whether the emotion attributes reach the standard or not, if yes, reasonably designing the training parameters of the detector, otherwise, retraining; taking the output of the image emotion detector and the convolution characteristics of the original image as the input of an emotion label classifier, and training classifier initialization parameters by adopting a gradient descent method; testing the prediction precision of the label classifier and judging whether the prediction precision reaches the standard or not, namely, reasonably designing training parameters of the label classifier when the prediction precision reaches the standard, otherwise, retraining; classifying the image emotion tags, and analyzing the image emotion. According to the method, the influence caused by a semantic gap can be reduced, image emotion prediction is more accurate, and the method is better suitable for large-scale image emotion classification tasks.

Description

technical field [0001] The present invention relates to the technical field of image processing methods, and more specifically, relates to an image emotion analysis method based on a multi-task learning method. Background technique [0002] Due to the urgent need for emotional expression awakened by visual content, scholars at home and abroad have begun to research and explore image sentiment analysis, but most of the existing research is mainly based on low-level visual features for sentiment analysis. There is a huge semantic gap in the labels, therefore, the existing image sentiment label classifiers are not ideal for image sentiment classification. Contents of the invention [0003] In order to overcome the problem of the huge semantic gap between low-level visual features and high-level emotional labels caused by the existing image sentiment analysis method, the present invention sets an objective function based on multi-task learning to train the emotional attribute ...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the problem of the huge semantic gap between low-level visual features and high-level emotional labels caused by the existing image sentiment analysis method, the present invention sets an objective function based on multi-task learning to train the emotional attribute detector, and applies the detected emotional attributes to In terms of visual features, the distribution of emotional labels and emotional attributes tends to be consistent, which reduces the impact of semantic gaps to a certain extent, makes image emotion prediction more accurate, and is well applicable to large-scale image emotion classification tasks. , providing an image sentiment analysis method based on multi-task learning

Images