Klasifikasi Ekspresi Wajah Menggunakan Convolutional Neural Networks (CNN) pada Dataset FER-2013

Abstract

Automatic facial expression recognition has become a rapidly growing research field driven by advancements in artificial intelligence and computer vision. However, facial expression classification still faces challenges, particularly in distinguishing expressions with similar characteristics. This study aims to develop a facial expression classification model using Convolutional Neural Networks (CNN) on the FER-2013 dataset. The research stages include data collection and preprocessing, CNN architecture design, model training using the Adam optimizer and categorical crossentropy loss function, and performance evaluation based on accuracy and the confusion matrix. The results indicate that the CNN model can recognize various facial expressions, achieving a maximum validation accuracy of 67.8%. Nevertheless, the model is still able to distinguish certain expressions accurately. Utilizing pretrained models such as VGG-16 or ResNet and implementing transfer learning techniques could enhance model accuracy and stability. With further development, this model has the potential to be applied in various fields, including facial expression-based security systems, human-computer interaction, and emotion analysis.