Pothole Detection Web App: an approach for detecting potholes in asphalt pavements using YOLO

Authors

  • Jean Cássio Peres Barbosa Universidade de Passo Fundo
  • Francisco Dalla Rosa Universidade de Passo Fundo
  • Rafael Rieder Universidade de Passo Fundo

DOI:

https://doi.org/10.5335/rbca.v16i3.15851

Keywords:

asphalt pavement, pothole detection, real-time object detection, web app, YOLOv7-tiny

Abstract

The detection of potholes in asphalt pavements is a crucial task for road safety. This survey requires significant time and financial resources. An automatic method for this task can help in Pavement Management System, speeding up processes of road recovery and maintenance of materials. With this in mind, this work presents the Pothole Detection Web App, a smart web application for detecting potholes in asphalt pavements using the YOLOv7-tiny architecture. The
solution allows performing the identification of potholes in asphalt pavements by photos, videos or live broadcasts.
The images or videos are sent to a web server, where the detection model is applied and the results are returned to the user. Preliminary tests pointed out good results, with an accuracy of 74% (F1-Score = 66%). The solution proved to be capable of detecting potholes and estimating defect dimensions with a good approximation, in reports with organized information. In addition, it showed good performance for real-time analysis, considering different browsers. The approach can serve both urban roads, using internet, and open roads, using communication with a local area network.

Downloads

Download data is not yet available.

Published

2024-12-03

Issue

Section

Original Paper

How to Cite

[1]
2024. Pothole Detection Web App: an approach for detecting potholes in asphalt pavements using YOLO. Brazilian Journal of Applied Computing. 16, 3 (Dec. 2024), 25–36. DOI:https://doi.org/10.5335/rbca.v16i3.15851.