Efficient YOLOv8 Model for Licence, Helmet and Person Detection

The personalized YOLOv8 model is optimized for security and surveillance automation with emphasis on commercial processing for applications such as person detection, helmet detection, and license plate reading. The accuracy and computational performance are enhanced by the incorporation of a Spatial Pyramid Pooling-Fast (SPPF) layer, depthwise separable convolutions, and a light MobileNet backbone. The customized YOLOv8 model attains a mean Average Precision (mAP) of 0.79 and a recall score of 0.93, outperforming the standard YOLOv8 and YOLOv5 models, as per experimental findings. The proposed system is highly efficient in detection and computation, thus being an effective solution for real-time security and surveillance in a range of applications, including small computing devices.