• Santosh E
• Mehwish
• Yashwanth N
• H Varshitha
• Tejas Manjunath

Potholes are a persistent and hazardous issue on roadways, contributing significantly to vehicle damage, traffic delays, and accidents—particularly during adverse weather conditions. Their unpredictable formation, varying in size and depth, poses serious challenges for timely detection and maintenance, making road infrastructure upkeep a major concern for authorities worldwide. Traditional pothole detection methods such as manual surveys, vibration-based sensors, or 3D laser scanners are either time-consuming, costly, or require complex setups and dedicated hardware. While recent advances in deep learning have introduced object detection models like YOLOv3, YOLOv5, and YOLOv7 for pothole identification, these approaches often lack depth estimation capabilities, leading to limited understanding of pothole severity and repair requirements. Furthermore, many existing solutions struggle with real-time performance and adaptability to diverse road conditions. To address these limitations, our project proposes an intelligent and cost-effective pothole detection system that combines the real-time object detection power of YOLOv8 with MiDaS v3.1-based monocular depth estimation. YOLOv8 enables precise and fast identification of potholes from road images or videos, while MiDaS allows estimation of real-world depth and area using a single image. This integration facilitates not only detection but also severity analysis and repair cost estimation, providing a holistic solution. Our system is trained on a custom-labeled dataset that includes varied road surfaces and lighting conditions, ensuring robust performance across environments. It is optimized for deployment on edge devices or in smart city systems, offering real-time detection with an average precision of 91.11% and inference time of 8.8 ms per frame. In conclusion, our project presents an efficient, scalable, and low-cost AI-based pothole detection system that overcomes the limitations of existing models by enabling both identification and quantification of road surface damage, paving the way for smarter road maintenance and safer transportation systems.

Potholes, YOLO, MiDaS, AI

"Automatic Pothole Detection System ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 5, page no.h329-h350, May-2025, Available :http://www.jetir.org/papers/JETIR2505831.pdf

"Automatic Pothole Detection System ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 5, page no. pph329-h350, May-2025, Available at : http://www.jetir.org/papers/JETIR2505831.pdf