Vehicle technology with autonomous systems is advancing rapidly, necessitating improved safety measures for road hazards such as potholes, which are a leading cause of vehicle accidents. Especially at night, human visual limitations highlight the importance of real-time pothole detection to enhance driver safety. Among object detection algorithms, the YOLO series stands out for its high speed and efficiency, making it suitable for real-time applications in dynamic environments. This study presents a detailed performance comparison of the YOLOv7 and YOLOv8 algorithms for detecting potholes on road surfaces using images from web-based sources and Google Earth. A comprehensive dataset of pothole images was constructed, enabling a robust evaluation of each model under various environmental conditions, such as lighting and weather variability. Key performance metrics, including precision, recall, and detection time, were analyzed to assess the strengths and weaknesses of YOLOv7 and YOLOv8, revealing significant differences in their accuracy and real-time capabilities. The findings provide insights into each algorithm’s efficacy and suitability for potential real-time implementation in autonomous vehicle safety systems, paving the way for future research on optimized models for road safety and maintenance applications.