<![CDATA[Background: Stress is a common mental health issue that requires early detection to prevent severe consequences.Aims: This study aims to develop an Android-based mobile application capable of detecting stress levels using facial expression analysis.Methods: The application was developed using the Rapid Application Development (RAD) approach. A Convolutional Neural Network (CNN) based on EfficientNetB0 was used and trained on the FER2013 dataset. The trained model was then converted into TensorFlow Lite format for efficient deployment on mobile devices.Results: The model achieved an accuracy of 80.23%, precision of 80.60%, recall of 73.84%, and F1-score of 77.07%. Furthermore, real-world performance testing on mobile devices demonstrated high efficiency. The optimized TensorFlow Lite model achieved a rapid inference time of 6.00 ms and a compact footprint of 2.85 MB. The developed application successfully performs real-time and offline stress detection using images captured from the camera or selected from the gallery. Functional testing using the black-box method showed that all application features operated correctly, while usability evaluation using the System Usability Scale (SUS) produced an average score of 79, indicating that the application is easy to use. Conclusion: This study demonstrates that integrating machine learning into mobile applications can provide a practical and accessible solution for early stress detection. However, dataset imbalance remains a limitation that affects model performance and should be addressed in future work.]]>
Copyrights © 2026
