Investigation on a vision-based approch for smart pothole detection using deep learning based on fast CNN

Alsharafi, Ashraf Khaled and Muhammed Nafis, Osman Zahid (2023) Investigation on a vision-based approch for smart pothole detection using deep learning based on fast CNN. Mekatronika - Journal of Intelligent Manufacturing & Mechatronics, 5 (2). pp. 87-99. ISSN 2637-0883. (Published)

Preview

Pdf
Investigation on a Vision-Based Approch For Smart Pothole Detection.pdf
Available under License Creative Commons Attribution Non-commercial.
Download (1MB) | Preview

DOI/Official URL: https://doi.org/10.15282/mekatronika.v5i2.9813

Abstract

The quality of road these days are important and roads always dangerous since its filled with potholes and damages which cause a lot of incident and numbers gets more increased in crowded area , this article investigates and compare the performance metrics of different object detection models that utilized the Fast CNN structure in it's backbones , Four processes make up the standard method of pothole detection: data acquisition, data pre-processing, feature extraction, and pothole classification. for the task of pothole detection. The study focuses on the evaluation of YOLOv6n, YOLOv8n, YOLOv5n, and YOLOv7 models using a dataset of road images containing pothole instances. The performance metrics analyzed include precision (P), recall (R), mean average precision at 50% IoU (mAP@.5), and mean average precision from 50% to 95% IoU (mAP@.5:.95) . The findings indicate that YOLOv8n demonstrates the highest overall performance, achieving significant precision and recall rates. These results provide valuable insights into the effectiveness of object detection models for pothole detection, contributing to the field of road maintenance and safety. The outcomes of this study can assist in the development of intelligent systems for automated pothole detection and maintenance planning

Item Type:	Article
Uncontrolled Keywords:	Yolo models; Mean Average Precision; Smart Pothole Detection; Vision-Based Approch
Subjects:	T Technology > TJ Mechanical engineering and machinery
Faculty/Division:	Institute of Postgraduate Studies Faculty of Manufacturing and Mechatronic Engineering Technology
Depositing User:	Mrs Norsaini Abdul Samat
Date Deposited:	18 Jan 2024 07:08
Last Modified:	18 Jan 2024 07:08
URI:	http://umpir.ump.edu.my/id/eprint/40089
Download Statistic:	View Download Statistics

Actions (login required)

View Item