Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{201355,
        author = {Vivek Kumar Singh and Manish Singh and Ram Anuj Gupta and Anurag Kumar and Ritesh Kumar Singh},
        title = {AI Based Pothole Detection System: An Intelligent Edge-Based System for Real-Time Road Damage Detection},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {5415-5421},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=201355},
        abstract = {Road infrastructure degradation continues to pose a significant challenge for governments and urban planners worldwide. Among the various types of road damage, potholes and surface cracks are the most dangerous, contributing directly to traffic accidents, vehicle damage, and rising maintenance costs. Traditional inspection methods, which rely on periodic physical surveys, are proving inadequate in the face of expanding road networks and growing urban populations. This paper introduces AI Based Pothole Detection, an intelligent, edge-optimized road damage detection system that brings automated monitoring capabilities to real-world deployment scenarios. The proposed system leverages MobileNetV2, a lightweight convolutional neural network architecture trained using transfer learning, to classify four distinct categories of road surface damage: longitudinal cracks, transverse cracks, alligator cracks, and potholes. The model was trained on the publicly available RDD2022 dataset and deployed through a full-stack Node.js and Edge Impulse pipeline, enabling real-time inference on edge devices with response latencies below 100 milliseconds. Experimental evaluation shows that the system achieves classification accuracy in the range of 85–92% across damage categories. Additionally, the system incorporates entropy-based uncertainty quantification, AI-driven cost estimation, severity assessment, and heatmap visualization features, which collectively provide a production-ready monitoring tool. The presented solution demonstrates that combining edge AI with transfer learning and a lightweight web interface creates a scalable, accessible, and efficient framework for road maintenance management, contributing to the goals of smart transportation systems and intelligent urban infrastructure.},
        keywords = {Road damage detection, deep learning, MobileNetV2, edge AI, transfer learning, convolutional neural network, smart city, pothole detection, intelligent transportation},
        month = {May},
        }