EveCamNet: A Fusion Framework of Event and RGB Camera Towards Detecting Drivable Area for Autonomous Vehicles

Jyoti M Satihal, Safa Imtihaz Sayyad, Ujwala Patil

2025

Abstract

We propose EveCamNet, a novel framework that fuses features from RGB and event cameras to detect drivable areas in dynamic scenarios. This framework enhances semantic segmentation for identifying drivable regions in autonomous driving systems. While RGB cameras excel at capturing spatial and textural details, they often struggle in dynamic conditions, such as motion blur and varying lighting. In contrast, event cameras offer high temporal resolution and low latency, but they lack detailed spatial context. Our proposed framework effectively combines the strengths of both modalities through attention-based feature fusion mechanisms and a robust loss function incorporating Cross-Entropy and Dice losses. Using an open-source dataset, we evaluate the performance of EveCamNet, achieving a mean Intersection over Union (mIoU) of 69.94% and Pixel Accuracy (PA) of 90.54%. These results highlight the potential of RGB-Event fusion as a promising approach for advancing autonomous driving systems.

Paper Citation

in Harvard Style

Satihal J., Sayyad S. and Patil U. (2025). EveCamNet: A Fusion Framework of Event and RGB Camera Towards Detecting Drivable Area for Autonomous Vehicles. In Proceedings of the 3rd International Conference on Futuristic Technology - Volume 2: INCOFT; ISBN 978-989-758-763-4, SciTePress, pages 135-143. DOI: 10.5220/0013588200004664

in Bibtex Style

@conference{incoft25,
author={Jyoti M Satihal and Safa Imtihaz Sayyad and Ujwala Patil},
title={EveCamNet: A Fusion Framework of Event and RGB Camera Towards Detecting Drivable Area for Autonomous Vehicles},
booktitle={Proceedings of the 3rd International Conference on Futuristic Technology - Volume 2: INCOFT},
year={2025},
pages={135-143},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013588200004664},
isbn={978-989-758-763-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 3rd International Conference on Futuristic Technology - Volume 2: INCOFT
TI - EveCamNet: A Fusion Framework of Event and RGB Camera Towards Detecting Drivable Area for Autonomous Vehicles
SN - 978-989-758-763-4
AU - Satihal J.
AU - Sayyad S.
AU - Patil U.
PY - 2025
SP - 135
EP - 143
DO - 10.5220/0013588200004664
PB - SciTePress