This research highlights the efficiency of YOLO-NAS and Detectron2 algorithms in detecting road defects like cracks and potholes using drone-captured aerial images. YOLONAS demonstrates significant accuracy, achieving a mAP score of 71.23% and F1 score of 70.04%, outperforming previous YOLO versions. Detectron2 exhibits an AP score of 55%, surpassing state-of-the-art experiments in coco instance segmentation. Both models display confidence values close to 100%, ensuring reliable object detection. The results show the potential of integrating drone-based inspection systems with deep learning algorithms to improve road safety, reduce manual efforts, and enhance infrastructure management. This approach can contribute to a country’s economic and social progress by facilitating efficient road maintenance and defect detection. Future implementations may involve real-time detection using drones for timely road defect assessment and decision-making.