<![CDATA[The rapid development of smartphone technology has increased the duration of device usage, which can lead to eye fatigue, decreased concentration, and drowsiness. Many users continue using smartphones despite experiencing signs of fatigue, potentially affecting health and productivity. Therefore, this study aims to develop a real-time drowsiness detection system for smartphone users based on Android using Google ML Kit and MobileNetV1. The system utilizes Google ML Kit for face detection and MobileNetV1 for classifying eye conditions into open and closed states. The dataset used is a combination of a public dataset from Kaggle (dataset_B_Eye_Images) and additional data collected independently to improve model generalization. The model was trained and further optimized through a fine-tuning process. Experimental results show that the model achieved an accuracy of approximately 96%, with balanced precision, recall, and F1-score values. The confusion matrix analysis indicates improved performance after fine-tuning. In real-time implementation, the system operates at 6.0–7.3 FPS with a latency of 60–72 ms per frame and a notification response time of less than 1 second. The system demonstrates robustness under varying lighting conditions, achieving accuracy up to 100% in bright conditions, 98% in normal conditions, and 95% in low-light conditions. However, performance decreases when users wear glasses due to reflection interference. Overall, the results indicate that MobileNetV1 is effective for real-time drowsiness detection on mobile devices, although further improvements are needed to enhance system robustness under diverse user conditions]]>
Copyrights © 2026
