Pothole Sensing and Depth Estimation System using Deep Learning Technique

Roads, serving as vital transportation routes, undergo wear due to heavy usage and environmental factors, necessitating regular maintenance. However, neglect or the impracticality of monitoring every location often results in the formation of potholes, contributing to traffic disruptions and acciden...

Full description

Saved in:
Bibliographic Details
Published in2024 OPJU International Technology Conference (OTCON) on Smart Computing for Innovation and Advancement in Industry 4.0 pp. 1 - 4
Main Authors Singh, Angad, Randhawa, Mannat, Kalra, Nidhi, Seth, Chandvi, Gill, Anhad, Chopra, Palika, Kaur, Gurnoor
Format Conference Proceeding
LanguageEnglish
Published IEEE 05.06.2024
Subjects
Accuracy
Computational modeling
Computer vision
Deep learning
detection
edge
Estimation
MiDaS
pothole
ResNet-50
Road safety
Technological innovation
Online AccessGet full text

Cover

More Information
Summary:Roads, serving as vital transportation routes, undergo wear due to heavy usage and environmental factors, necessitating regular maintenance. However, neglect or the impracticality of monitoring every location often results in the formation of potholes, contributing to traffic disruptions and accidents. This paper explores a real-time pothole detection and depth estimation system employing computer vision and deep learning models. Utilizing image processing techniques, the system algorithm capable of identifying potholes and estimating their depths using a smartphone's camera and an ultrasonic sensor. Implemented using a PyTorch library in Python, the system employs a fine-tuned ResNet-50 architecture on a specialized pothole classification dataset, training it effectively for pothole identification. Additionally, the MIDAS model, enables precise distance calculations for both potholes and their surroundings, thereby facilitating accurate depth estimation.
DOI:10.1109/OTCON60325.2024.10687877