Emotional Analysis with Deep Learning

Abstract

– Image processing, one of the most exciting developments in deep learning, has made significant progress in various fields today. Suspicious behavior detection, social media user sentiment analysis, customer feedback, marketing and sales strategies, and many other areas can now analyze human reactions. Studies on the analysis of facial expressions date back to the 19th century. In 1872, Darwin proposed the universality of facial expressions. Therefore, dealing with facial expressions is not a new concept. Based on this, this project aims to integrate the concept of emotion analysis with the innovations of the modern era. According to a study conducted in 2012, Convolutional Neural Networks (CNN) are the most successful deep learning algorithm for object recognition. CNNs can distinguish features in images using numerous hidden layers. Filters that operate on matrices are used in each layer. These filters are the building blocks of the feature extraction process and play a crucial role in recognizing features such as edges, lines, colors, and other visual patterns. In this paper, a CNN model was built using the Python programming language and the Keras library. The model was trained using the FER-2013 dataset, and images of facial expressions were classified within the dataset. The performance of the model was evaluated using loss and validation metrics. Optimization efforts were made on the model to observe the changes in validation metrics due to training parameters. With an optimized model based on this information, simultaneous emotion analysis, age estimation, and gender prediction can be performed.