Facial Emotion Recognition (FER) using Convolutional Neural Network (CNN)

Abstract

Our research tackles the challenge of real-time facial emotion recognition (FER) from live video feeds, a crucial problem in various fields like security access, customer satisfaction analysis, and mental health diagnosis. Unlike previous studies focused on static images, our approach provides immediate emotion predictions directly from video streams. We implement Deep Convolutional Neural Networks (DCNN) for facial image classification. Moreover, we used pre-trained models like EfficientNet, ResNet, VGGNet and a Haar face classifier to achieve an impressive 82% accuracy on the training data of FER2013 dataset. This technology holds promise for human-computer interaction and mental health applications. However, it also raises ethical concerns, emphasizing the need for responsible deployment. Our work addresses the practical limitations of existing FER methods and bridges the gap between laboratory research and real-world applications, offering a valuable contribution to the field.