CATTLE DETECTION AND MISHAP AVOIDANCE SYSTEM USING YOLO v5 ALGORITHM
DOI:
https://doi.org/10.58973/CTRJ.22115Keywords:
Object detection, Road Safety, CNN, YOLO, , Mishap AvoidanceAbstract
Background: Stray cattle on Indian roads pose a significant road-safety hazard, and vehicles lack a reliable, low-cost way to detect them and prevent collisions.
Objective: This study proposes a robust vehicular cattle detection model using YOLO version 5 to help vehicles avoid mishaps on Indian roads.
Methods: A custom dataset of more than 500 annotated on-road cattle images was used to train and validate a YOLOv5-based single-stage object detector, evaluated across training epochs from 10 to 150 and tested on real-world video samples.
Results: The model achieved 90.5% average accuracy, 81.5% average sensitivity, and 89.5% average F-measure, with true-positive detection of 82-85% on test video samples.
Conclusion: The proposed YOLOv5-based cattle detection model offers a practical, effective approach to reducing vehicle-animal collisions on Indian roads.
Keywords: Object detection, Road Safety, CNN, YOLO, Mishap Avoidanc
HIGHLIGHTS:
1. YOLO v5 is used for developing a robust cattle detection model.
2. Real world video samples and on-road cattle image datasets showed 90.5% accuracy.
Downloads
Published
Issue
Section
License
Copyright (c) 2022 Computing Technology Research JournalAll articles published in the Computing Technology Research Journal (CTRJ) are freely available through the journal website for personal, educational, and research purposes. Readers may read, download, copy, print, cite, and share published articles for non-commercial use, provided that appropriate credit is given to the original authors and the journal.
Commercial reproduction, redistribution, systematic storage, translation, adaptation, or use of the published content for commercial purposes is prohibited without the prior written permission of Sevas Publishing.
Authors are responsible for obtaining permission to reproduce any copyrighted material included in their manuscripts and for acknowledging the original source where required.
