Emotion Recognition from Facial Expressions Using Deep Learning Model

Abstract

Facial expression recognition is a crucial research area in computer vision with applications in human-computer interaction and affective computing. This study uses the KDEF and PICS datasets to develop an emotion recognition application leveraging deep learning techniques. A new model was created using Convolutional Neural Networks (CNNs) for feature extraction. Deep learning models require extensive datasets for optimal performance. The KDEF dataset contains 4,900 images, while the PICS dataset has only 322 images. To address this limitation, data augmentation techniques were applied, expanding the PICS dataset to 4,830 images. Both datasets were then trained separately. The model classifies seven emotions: fear, anger, disgust, happiness, neutral, sadness, and surprise. It achieved 97.44% accuracy on the KDEF validation set and 98.24% accuracy on the PICS validation set. © 2024 IEEE.