Drowsiness is a sign that the body is entering a phase towards rest. Integrated detection of sleepy activity light control where light control automation is carried out based on the results of object detection, namely humans when they are sleepy, will be needed to efficiently use electrical energy. Analysis of the sleep detection scheme and the effectiveness of the object detection method in detecting sleepiness with body and eye position parameters was carried out using the Efficientdet Lite object detection method as the architecture used in making sleep detection models where the model created is implemented on a Raspberry Pi which is integrated with light control resulting in detection drowsiness with a comparison of 3 sleepiness parameters (body position, eyes, facial expressions) is able to control lights in the form of automatic lights turning off and on based on the labels that appear in each drowsiness detection parameter class. The results obtained were that the metric evaluation produced a different Average Precision for 3 parameters but in the appearance of the bounding box the most precise in detecting objects was the sleeping position, for the eye parameter the bounding box was not appropriate in detecting drowsiness while in facial expressions the bounding box was difficult to appear. Accuracy testing was carried out on the sleep position parameter resulting in a detection accuracy of 80%.
Copyrights © 2023