Smart Pothole Detection and Road Monitoring Systems: A Review of Sensing Methods, Deep-Learning Detectors and an Implemented Repair-Tracking Prototype

Automated pothole detection can help road agencies identify damage, but a detection alone does not ensure a repair. This paper reviews 17 studies from 2020 to 2026. It covers smartphone sensing, camera-based detection, deep-learning models, public datasets, edge devices, poor-light conditions, severity estimation, and repair monitoring. Across the studies, reported accuracy and speed vary with the dataset, image conditions, hardware, and evaluation method. Their scores are therefore not directly comparable. A second finding is that many studies focus on detecting or mapping damage, while fewer describe the full path from a report to contractor assignment and repair verification. We also describe our web prototype, which accepts citizen and dashcam reports, matches them to registered roads, supports contractor updates, stores repair photos, and allows an administrator to review repairs. The prototype shows how these tasks can be connected in one workflow. We outline tests using RDD2022 and local images, including model metrics, weather and lighting groups, surveyed GPS points, and edge hardware.