A Deep Learning Model for Visual Sentiment Analysis of Social Media

Visual sentiment analysis seeks to comprehend the emotional responses that images elicit in viewers. Despite the fact that this topic is relatively new, a wide variety of strategies have been created for diverse data sources and issues, leading to a substantial body of study. This Thesis attempts to provide a thorough overview of the subject by reviewing key literature. The topic is covered under many primary headings after a description of the task and the relevant applications. The article describes the design principles of broad Visual Sentiment Analysis systems from the three main perspectives of emotional models, dataset specification, and feature design. Recent Deep FER systems have typically concentrated on two main problems: overfitting brought on by a lack of training data and expression-unrelated factors including illumination, head posture, and identification bias. Authors present a full evaluation of Deep FER in this study. This study presents details on FER dataset with DEEP Learning model i.e. CNN and comparative analysis with other models such as SVM, K-Means. Convolutional neural networks (CNNs), in particular, have among all FER techniques demonstrated enormous potential because to their robust automated feature extraction and computational efficiency. In this work, we achieve the highest classification accuracy on the FER2013 dataset.