YOLOv8 for Pedestrian Detection: A Comparative Study for Pedestrian Detection
Sivaraman G, E. Sophiya, M. Diviya · 2024
This research analyzes the usefulness of YOLOv8 for robust pedestrian recognition in demanding real world situations. A custom-trained YOLOv8 model is evaluated against SSD models, displaying better accuracy in detecting and localizing human classes within images characterized by occlusions and variable scales. We analyze the model’s performance across confidence thresholds, demonstrating a positive balance between precision and recall. While noting computational considerations, our findings demonstrate YOLOv8’s potential for advancing pedestrian detection systems, opening the way for enhanced safety and efficiency in applications such as autonomous driving, surveillance, and smart city efforts.