YOLOv8-EVdet: An emergency vehicle detection method for autonomous vehicles
Abstract
This work presents a deep-learning based approach for emergency vehicle detection and classification for various driving conditions. Conventional vision-based methods often struggle with poor lighting, motion blur, and occlusions caused by surrounding vehicles, leading to inaccurate detections that may eventually slow the emergency vehicles' response. To address these challenges, we propose a modified object detection algorithm, tailored to identify emergency vehicles reliably in varied traffic and environmental conditions. Since existing datasets provide limited coverage of emergency vehicle classes, we additionally provide a custom dataset that includes diverse emergency vehicle classes in challenging visual and traffic scenarios. The training process involved extensive data augmentation and fine-tuning, to ensure robustness against occlusion, low lighting, and diverse backgrounds. The proposed method was evaluated on the custom EVdet dataset achieving a mAP@50 of 0.782 ± 0.006, across three independent training runs, confirming training stability. Cross-dataset evaluation on nuImages and EMS3D-KITTI demonstrated generalization to unseen domains. Additionally, a zero-shot comparison against four COCO-pretrained baselines (YOLOv5m, YOLOv8m, YOLO11m, RT-DETR-L) revealed that general-purpose detectors score 0.0 mAP on all emergency vehicle classes, establishing the necessity of task-specific fine-tuning; paired bootstrap and Wilcoxon signed-rank tests confirm that the improvements over all baselines are statistically significant (p < 0.01).
Related
- Conditional social acceptance of electric vehicles and V2G across transition pathways in a scenario based study
- On-road emissions of light-duty vehicles: detailed analysis of particulate and gaseous pollutants
- Investigation of VOC emission characteristics of light-duty gasoline vehicles and combustion-side formation mechanisms of key VOC species
- Adsorptive removal of benzo(ghi)perylene from vehicle-wash wastewater using eucalyptus wood waste-derived biochar
- High-resolution 3D turbulence measurements with different quadcopter uncrewed aerial systems: verification and validation for single- and multi-vehicle measurements
BibTeX
@article{loukas2026yolov8evdet,
title = {{YOLOv8-EVdet: An emergency vehicle detection method for autonomous vehicles}},
author = {Petros Loukas and David Bassir and Angelos Amanatiadis},
journal = {International Journal of Advanced Robotic Systems},
year = {2026},
month = sep,
doi = {10.1177/17298806261486243},
url = {https://hal.science/hal-05739965v1},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - YOLOv8-EVdet: An emergency vehicle detection method for autonomous vehicles
AU - Petros Loukas
AU - David Bassir
AU - Angelos Amanatiadis
PY - 2026
DA - 2026/09/07
JO - International Journal of Advanced Robotic Systems
DO - 10.1177/17298806261486243
UR - https://hal.science/hal-05739965v1
AB - This work presents a deep-learning based approach for emergency vehicle detection and classification for various driving conditions. Conventional vision-based methods often struggle with poor lighting, motion blur, and occlusions caused by surrounding vehicles, leading to inaccurate detections that may eventually slow the emergency vehicles' response. To address these challenges, we propose a modified object detection algorithm, tailored to identify emergency vehicles reliably in varied traffic and environmental conditions. Since existing datasets provide limited coverage of emergency vehicle classes, we additionally provide a custom dataset that includes diverse emergency vehicle classes in challenging visual and traffic scenarios. The training process involved extensive data augmentation and fine-tuning, to ensure robustness against occlusion, low lighting, and diverse backgrounds. The proposed method was evaluated on the custom EVdet dataset achieving a mAP@50 of 0.782 ± 0.006, across three independent training runs, confirming training stability. Cross-dataset evaluation on nuImages and EMS3D-KITTI demonstrated generalization to unseen domains. Additionally, a zero-shot comparison against four COCO-pretrained baselines (YOLOv5m, YOLOv8m, YOLO11m, RT-DETR-L) revealed that general-purpose detectors score 0.0 mAP on all emergency vehicle classes, establishing the necessity of task-specific fine-tuning; paired bootstrap and Wilcoxon signed-rank tests confirm that the improvements over all baselines are statistically significant (p < 0.01).
SN - 1729-8806
SN - 1729-8814
ER -