<![CDATA[Recognizing human facial expressions has broad benefits in various fields. For example, in the field of psychology, by analyzing a person's facial expressions during the counseling process, a psychologist can understand a patient's emotional changes and identify psychological problems. One of the popular algorithms for facial expression recognition is the Convolutional Neural Network (CNN). In this study, an architectural model of the Convolutional Neural Network (CNN) is used which consists of three convolution layers. The test results show that the model drilled with ADAM optimization, batch size 32, and data augmentation achieved good accuracy, namely 70.16% for training data and 64.43% for data validation at the 100th epoch. This study also conducted tests using facial expression images from self-made datasets and achieved the highest accuracy of 67% after training the model up to the 100th epoch. The program we created succeeded in recognizing facial expressions well in real-time situations in 20 participants of various ages. However, this study shows several improvements that can be made, such as increasing the quality and quantity of facial expression data and developing the CNN model with additional features to improve accuracy and overcome overfitting.]]>
Copyrights © 2023
