MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641100341 A) filed by Vignans Foundation For Science, Technology And Research on August 19, 2026, for An Intelligent Deep Learning Framework For Automated Pothole Detection.
Inventors include Dr. K. Santhi Sri; Mr. Irfan Sayyad; Mr. Peddipaga Rahul; Mr. K. Raghavendra Murali; and Ms. T. Sri Vyshnavi.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: ABSTRACT [0016] Road surface deterioration caused by potholes remains a widespread problem that leads to vehicle damage, traffic delays and accidents. Conventional inspection methods rely on manual surveys or specialised vehicles and cannot provide continuous coverage of large road networks. The present invention discloses an intelligent deep learning framework for automated pothole detection that integrates YOLOv5 for real-time localisation of potholes and ResNet-18 for binary verification of their presence. Road images first undergo pre-processing that includes resizing, normalisation, annotation conversion and data augmentation. The YOLOv5 model generates bounding boxes that locate candidate potholes with high spatial accuracy, while the ResNet-18 model independently classifies the same images to confirm or reject the detections, thereby reducing false positives arising from shadows, road markings or surface irregularities. The combined output supplies reliable pothole information that can be reported automatically and later linked to GPS mapping systems. Experimental evaluation shows high detection accuracy and real-time performance, enabling lower inspection costs, minimal human intervention and earlier maintenance actions that contribute to safer transportation infrastructure.
Disclaimer: Curated by HT Syndication.
