Ines Bessrour
Apr 2023
Paper
Abstract : Our primary focus and dedication revolved around the development of a sentiment analysis interface using advanced computer vision techniques. This project focuses on the development of a deep learning model using convolu- tional neural networks (CNNs) for emotion detection from facial expressions. The aim of this project is to accurately classify facial expressions into seven categories: Angry, Disgust, Fear, Happy, Sad, Surprise, and Neutral. We utilize the FER-2013 dataset, a widely used benchmark dataset, for training our model. The project begins with a comprehensive literature review, exploring various approaches in emotion classification. Our proposed solution involves designing a CNN model that extracts discriminative features from facial images. The model architecture consists of convolutional, pool- ing, and fully connected layers. Additionally, we incorporate data preprocessing techniques to enhance the model’s performance. Implementation of the solution involves training the model on the FER-2013 dataset, fine-tuning hyperparameters, and evaluating the model’s performance on a separate test dataset. Our findings demonstrate promising results in accurately classifying facial expressions into different emotions.