Evaluating YOLOv8 and YOLOv11 for Real-Time Person Detection in Crowded Environments

Surveillance, autonomous operations, and public safety rely heavily on people detection. Conventional means of detection by human efforts mostly include inefficiency, time consumption, and bias. This makes deep learning techniques a must for the sake of automation. The study presented in this paper deals with the performance of two deep learning models, namely YOLOv11 and YOLOv8, in person detection on a benchmark dataset.YOLOv11, with its architectural improvements, presents with more accuracy: it maintains accuracy despite occlusions in complex scenes. Improved detection performance is due to its enhanced feature extraction. Meanwhile, being one of the single-stage architectures, YOLOv8 has good inference speed beneath those specific conditions, making it capable of being adopted for real-time applications. The paper pinpoints the dilemma between mAP and precision, stressing the need to make the model choice depending on the dataset traits, computational resources, and practical use. The comparison provided by the experimental findings gives important insight. For the optimization of person detection systems via state-of-the-art deep learning architectures, this paper presents a comparative framework.