Jejak Kebahagiaan: Analisis Ekspresi Wajah Berbasis Image Processing untuk Mengukur Kebahagiaan Komunitas

Abstract

This research aims to analyze facial expressions using image processing techniques based on Convolutional Neural Network (CNN) and TensorFlow as a method to measure the happiness levels of a community. The dataset used is the affectnet-training-data from Kaggle, consisting of 96x96 pixel RGB images of faces, divided into training and testing sets to categorize faces based on displayed emotions. The CNN model developed in this study uses the ResNet-5 architecture, consisting of five residual blocks to address the vanishing gradient problem and enable deeper training. The training process involves iterating through the training data, calculating loss, and adjusting parameters using an optimizer to achieve optimal accuracy. The results show that the CNN model can recognize facial expressions with a high accuracy rate of 93%. These findings indicate that the proposed method can be used to measure the happiness of a community more objectively and measurably, providing new insights into the development of reliable and effective happiness measurement tools.