<![CDATA[This study discusses the development of a legitimate occupant detection system in smart rooms using the YOLOv8 algorithm, tested under daytime and nighttime lighting conditions. The system is designed using a Raspberry Pi connected to a webcam for real-time monitoring. The aim of this study is to evaluate the system's performance under different light intensities. Data were obtained by capturing images during the day and night, which were then used as a training dataset for the YOLOv8 model. With a mAP@0.5 of 0.91 and precision, recall, and F1-score values of 0.90, 0.88, and 0.89, respectively, the evaluation findings demonstrate that the system operates effectively under ideal lighting conditions. This shows that the model can recognize things accurately and consistently in real time. However, performance drastically declines in low light, with mAP@0.5 falling to 0.68 and precision, recall, and F1-score falling to 0.70, 0.65, and 0.67, respectively. This indicates a rise in false and missed detections (FP and FN). Reduced image quality, including inadequate illumination, noise, and poor feature visibility, is the primary cause of this degradation. However, it has been demonstrated that using more light sources increases detection accuracy]]>
Copyrights © 2026
