Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:T2-5
Session:
Number:4872
Exploring Visual Features of Emotional Images on Social Media
Satoshi Sanjo, Marie Katsurai ,
pp.555-558
Publication Date:2016/7/10
Online ISSN:2188-5079
DOI:10.34385/proc.61.4872
PDF download (6.7MB)
Summary:
In this paper, we explore visual characteristics of the emotional images on social media through automatic image generation using a class of Convolutional Neural Networks (CNNs) called deep convolutional generative adversarial networks (DCGANs). Results of experiments present the visual appearance of each emotion, from which we can observe its specific characteristics. Furthermore, based on the trained DCGANs, we investigate the similarity between public image datasets. Our findings are also consistent with sentiment classification across different datasets.