Design of an efficient visual sentiment Classification system using deep learning Approaches

Abstract

Psychological studies reveal that the human emotional response varies with the type of stimulus. Images are effective tools to convey rich semantics and to provoke strong emotions within an individual. With the explosive growth of social networks such as Twitter, Facebook and Flickr, individuals tend to express their emotions online by posting images, text, audio, and video or in any other form every day. This results in a huge amount of online data which in turn necessitates the comprehension of these visual contents. Emotion recognition is the key to understand human computer interaction. Image sentiment analysis or visual sentiment analysis is a new and promising research field that deals with the automatic detection of human emotions expressed in viewing images. It requires understanding of the high-level abstraction of the visual content. The advent of Artificial Intelligence has enabled machines to interpret images from a human perspective. Emotional intelligence is an interdisciplinary area of research spanning computer vision to psychology. newlineVisual sentiment analysis has emerged as a budding research area due to its great interest for emotions used in health care, education, entertainment, advertising, psychological and cognitive studies, journalism and image captioning. The sentiment characterizations can be used to tag the visual content. Automated tagging of images using sentiments is useful in recommendation and retrieval systems. Compared to other computer vision tasks, designing a sentiment model for a machine to recognize and interpret human emotions is quite a challenging task due to the affective gap, complexity and subjective nature of human emotions newline