Despite advancements in transportation technology, road-related accidents remain a significant global concern. The World Health Organization (WHO) reports approximately 1.24 million road accidents each year. These accidents often result from various obstacles or anomalies on the road, such as vehicles, pedestrians, potholes, and road damage. Detecting these road anomalies is crucial for reducing damage and improving overall traffic safety. Roads are a common mode of travel but are fraught with obstacles that can jeopardize individual safety, damage assets, and harm others. This research focuses on road anomaly detection using YOLOv8 and supervised machine learning models. The paper targets anomalies such as light motor vehicles, heavy motor vehicles, pedestrians, road damage, speed bumps, potholes, and accidents. The YOLO (You Only Look Once) algorithm excels in real-time object detection, enabling the identification of anomalies as they occur. In addition to YOLO, other machine learning models like SVM (Support Vector Machine), Random Forest, Naive Bayes, and K-NN (K-Nearest Neighbors) (73%, 69%, 56%, 69% accuracy respectively) were utilized for classification tasks. YOLO models (mAP of 96%, 79% and 90% for Vehicles, Road Damages and Pedestrians respectively) are used to detect and annotate anomalies, while the classification models determine if an image contains accidents or other specific anomalies. Further, video frames from CCTV cameras were used for classification and anomaly detection. YOLO models detect various road anomalies, while machine learning models classify the presence of accidents and other types of road obstacles. The proposed models are highly replicable and can be adapted into various domains such as autonomous driving, traffic management, and urban planning, in these scenarios real-time detection and classification of obstacles are crucial for safety and efficiency.