In this study, a novel approach for defect detection and classification of road defects, such as cracks, potholes, and uneven surfaces, is presented using deep learning (DL) and computer vision (CV) techniques. The proposed method addresses limitations of manual road inspection, including human error and the tendency to overlook significant surface defects, which directly contribute to road hazards and infrastructure degradation. A key feature of this work is the use of the YOLOv7 instance segmentation (YOLOv7-seg) model for fast, real-time road condition monitoring, owing to its high efficiency and precision in detecting complex surface deformities. The footage was captured using a GoPro HERO 9 mounted on a moving vehicle traveling the Jamshoro-Hyderabad route, with further augmentation to ensure diversity and generalization of road defect detection. The model was fine-tuned and evaluated using precision and recall metrics, achieving 98.6% accuracy and 99% mean Average Precision (mAP), demonstrating high accuracy in defect detection. This approach can be scaled up to handle multiple sensors in a large infrastructure monitoring system, facilitating automated maintenance and enhancing road safety. Extending this system into IoT and smart city technologies could support advanced industrial infrastructure management in the future.