Automatic Visual Sentiment Analysis with Convolution Neural Network

There is strong demand for the application of automated sentiment analysis to visual and text contents in today’s digital world so as to significantly reveal people’s feelings, opinions, and emotions through texts, images, and videos in popular social networks. However, conventional visual sentimental analysis has been subject to some drawbacks including low accuracy and difficulty to detect people’s opinions. In addition, a considerable number of images generated and uploaded every day across the world complicate the already given problem. This paper aims to resolve the problems of visual sentiment analysis using deep-learning Convolution Neural Network (CNN) and Affective Regions (ARs) approach to achieve comprehensible sentiment reports with high accuracy.