Roads play a vital role in our daily lives, serving as primary fundamental component of the urban infrastructure. Daily transportation activities are heavily dependent on the availability and condition of the road infrastructure. Especially in metropolitan areas, people often face challenges due to deterioration of road conditions. These problems become more severe during the rainy season, as potholes fill with water, open manholes pose serious hazards, road surfaces develop various types of cracks, and debris accumulate, all of which compromise safety and traffic flow. We propose a system to assess road conditions and detect road anomalies, including potholes, waterlogged potholes, debris, cracks, and open manholes using state-of-the-art deep learning techniques. The proposed system employs object detection models based on the You Only Look Once (YOLO) framework, specifically YOLOv8. The results demonstrate steadily increasing performance in training and validation sets, with high precision and recall achieved for multiple classes of anomalies. The proposed system integrates a Flask-based back-end API with a user-friendly web interface to support real-time detection of static images, recorded videos, and live camera streams with minimal latency. We have collected a total of 5,148 images that capture different types of road anomalies under different road conditions on the road surface. Dataset is divided into training, testing, and validation sets. The proposed system uses a mean Average Precision at 50% (mAP_50) intersection-over-union (IoU) to measure the performance of a model. Our system achieved mAP_0.5 of YOLOv8-nano is 66.8%.