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{210197,
        author = {Dhruva Suresh and K Nandan and Kavana D S and Pavani B and Mrs. Ramya HJ and Dr krishna Kumar P R},
        title = {DriveSafe Vision: Real-time pothole detection using deep learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {6},
        pages = {38-46},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=210197},
        abstract = {Potholes endanger Indian road users, yet reporting them and tracking repair is slow and rarely accountable. DriveSafe Vision is a low-cost, responsive web application combining image-based pothole detection with geotagged crowdsourced reporting and deadline-based accountability. A server-side deep-learning module localises each pothole in a citizen's photograph, estimates its size and grades severity on four levels; the report is routed to the responsible area administrator with a fix deadline (seven days by default, configurable by area and severity). Overdue complaints trigger alerts and lower the administrator's on-time rate, which is combined with citizen ratings to classify performance. Detection runs zero-shot on a general-purpose multimodal transformer model; as a trained drop-in replacement, a fine-tuned YOLOv8n detector reached an [email protected] of 79.4 per cent on the held-out test split of a public 665-image pothole dataset. A functional pilot with eight test reports exercised routing, deadlines and escalation.},
        keywords = {Pothole detection; Object detection; Road damage; Crowdsourcing; Smart city; Service-level accountability},
        month = {November},
        }