<![CDATA[Efficient classroom management is a crucial requirement in academic environments such as the Faculty of Computer Science and Information Technology to increase productivity. This study aims to design and evaluate a real-time presence detection and counting system by implementing the Region of Interest (ROI) method to improve computational efficiency and accuracy. This methodology involves the use of a Logitech C270 HD webcam, with a static ROI set at 90% of the central video frame to focus the analysis. Person detection and counting are performed using a combination of Histogram of Oriented Gradients (HOG) for the body and Haar Cascade for the face. Time series reasoning with a minimum duration of 60 seconds and a grace period of 5 seconds is implemented to validate presence and stabilize the room status, with system performance evaluated using Precision and Recall metrics. The results show that the system successfully displays the status and number of people in the room very well, but the evaluation shows a Recall value of 1.00, which means the system detects every actual human presence. However, this system has significant accuracy issues, indicated by a low Precision of 0.04 and a high number of False Positives of 710. In conclusion, although the ROI application successfully improves the computational load and the temporal logic stabilizes the output, the HOG and Haar Cascade models are inadequate to handle visual noise in the ROI, resulting in low Precision and indicating the need for more sophisticated detection models.]]>
