Efficient human presence detection within specified Regions of Interest (ROIs) is essential to many real-world applications, including as resource allocation, security surveillance, and crowd monitoring. In this paper, we present a robust human pres-ence detection system leveraging advanced techniques, notably YOLOv8, for accurate and real-time object detection. Our system addresses the challenges associated with traditional methods by focusing on multi-scale detection windows and efficient merging of nearby detections to refine results. We introduce a streamlined approach for ROI definition and utilize transfer learning to adapt YOLOv8 to specific human presence detection tasks. Extensive experiments on diverse datasets demonstrate the effectiveness and robustness of our system, showcasing superior performance in terms of accuracy, real-time processing, and adaptability to chal-lenging scenarios. Additionally, we implemented four algorithms, including Histogram of Oriented Gradients (HOG) [1], Roboflow 3.0 object detection algorithm, YOLOv5, and YOLOv8. After thorough examination, we determined that YOLOv8 showed the most favorable outcomes in terms of both precision and speed. Hence, we opted for YOLOv8 as the primary algorithm for our system aimed at detecting human presence. This decision was informed by its superior performance across diverse datasets and its capability to address the challenges associated with traditional methods.