<![CDATA[Facial emotion detection is one of the essential technologies in human–computer interaction. This study aims to compare the performance of the Convolutional Neural Network (CNN) and the CNN–Long Short-Term Memory (CNN-LSTM) algorithms in detecting emotions using three datasets: FER2013, CK+, and AffectNet. The research process involved data collection, preprocessing, model training with transfer learning, and evaluation using accuracy, precision, recall, F1-score, and confusion matrix metrics. The results show that CNN achieved only 60% accuracy with varying precision and recall, whereas CNN-LSTM reached an accuracy of 80–87% with more stable performance. Analysis of accuracy curves, loss, log loss, and radar charts indicates that CNN-LSTM outperforms CNN in classifying emotions more evenly and consistently, although it requires longer computational time. These findings emphasize that integrating CNN and LSTM can enhance the effectiveness of facial emotion detection systems, particularly in handling complex and dynamic expressions.]]>
Copyrights © 2025
