YOLOv8- Driven Object Detection Framework for Real-Time Vehicles and Pedestrian Monitoring in Adverse Conditions

Abstract

The given research paper singles out one vehicle system that contributes to increasing the safety level by delivering instantaneous object detection and forecasting on the basis of YOLOv8. Unlike standard external surveillance systems, this one is enclosed within the car so as to be able to detect and classify pedestrians, motor vehicles, and road debris. Moreover, the technology can be fine-tuned to harsh weather conditions, including rain, fog and low light so that the technology will be able to operate safely and with a high degree of strength under various driving conditions. Using cutting edge visual recognition and deep learning networks, this technology can evaluate the potential of objects under observation to cause possible collisions and hazards by processing live stream video streams and, in turn, enable the vehicle to come up with safety-enhancing actions. The YOLOv8 model consumed fast and precise analysis with the help of the detected objects and provided significant contextual data of the surrounding of the vehicle. The model was compared to Precision, Mean Average Precision (mAP) and Intersection over Union (IoU) metrics, which proved the superiority of YOLOv8 models over previously developed object detection models regarding its ability to identify objects and make predictions about the danger. The model also undertook predictive risk analysis that was done with 643 high hazard image and 163 low hazard image and this displays the capability of the model to analyze safety situation in instantaneous. Situational awareness, which is a vital element of autonomous driving, will be improved with the facilitation of real-time applications.