Sentiment Analysis from Social Media Image using CNN

Abstract

Images are used on social media to convey views, opinions, feelings, and emotions. We research the issue of knowing natural feelings from large-scale pictures of social media in this work. We need to understand public sentiment for Social marketing, to develop product quality, to improve customer service, social media monitoring and research purpose. In this work we have classified happy, sad and angry emotions of people by using their face image on social media. For the classification of pictures, we have used Convolutional Neural Networks (CNN). First, we modified an appropriate CNN architecture for the assessment of picture sensitivity. The findings indicate that in picture sentiment analysis, the suggested CNN can attain stronger efficiency than rival algorithms. By using this model we can classify image into three categories of sentiment named happy, sad and angry and we got an accuracy of 75.28%. This work can be further enhanced to be used in many other related areas of image classification.