VGuard: Development of Mobile Real-Time Vehicle Damage Detection Application with Image Recognition using YOLOv8
Year: 2024
Type: conference-paper
Abstract: Innovations in the modern automotive industry have transformed vehicle maintenance, but inspections remain to be labor-intensive and lack visual representation. This study proposes VGuard, a mobile application that integrates On-Board Diagnostics (OBD-II) data retrieval and You Only Look Once-based (YOLOv8) image recognition for external damage detection, and a dual visualization of 2D and 3D to streamline vehicle health inspections. The app utilizes an ELM 327 scanner for engine data, while a dataset of 4,609 images of vehicle damage, specifically of rust, scratches, dents, and cracks, were used to train the image recognition model. This achieved an mAP of 72.1%, 82.7% precision, and 63.7% recall. Developed with Flutter, VGuard was evaluated by sixty (60) respondents using a 4-point Likert scale survey based on ISO 25010 standards. The app received an overall rating of 3.36, corresponding to “Excellent” and demonstrating its effectiveness in data retrieval, damage detection, and display. Recommendations include improvements in data dependency, detection accuracy, usability, user interface, and additional features for further development.
Institution: University of the City of Manila
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412119671