Pengembangan Aplikasi Deteksi Tingkat Stres Melalui Ekspresi Wajah Menggunakan Convolutional Neural Network Dan Tensorflow

Abstract

Stress is an emotional response that often arises due to psychological pressure, and if not handled properly can have a negative impact on physical and mental health. This research aims to develop an Android application that can detect user stress levels based on facial expressions using the Convolutional Neural Network (CNN) artificial intelligence model integrated with TensorFlow. Application development is carried out using the Modified Waterfall method, which is a software development method consisting of the stages of requirements analysis, system design, implementation, testing, and evaluation. The developed application uses the device's camera to capture images of the user's face, then processes them through a CNN model that has been converted into TensorFlow format. This model is able to classify facial expressions into four stress levels, namely no stress, weak stress, mid stress, and high stress. After detection, the system displays the result label and provides stress management advice according to the user's condition. The user authentication feature was developed using Firebase Authentication for login and account registration. Although data storage to Firebase Realtime Database has not been actively implemented, its configuration and dependencies have been prepared for further development. The test results show that the application runs well, is able to perform classification quickly, and provides useful feedback directly to the user.