HyEff-UNet: an attention-driven Efficientnet U-Net with hybrid loss optimization for UAV-based marine oil spill detection and segmentation
- 1chool of Computer Science and Engineering, Vellore Institute of Technology–Chennai Campus, Chennai, India
Abstract
IntroductionMonitoring oil spills in the port area is very difficult and can spread rapidly, resulting in considerable damage to the environment. This study proposes HyEff-UNet, a deep learning-based semantic segmentation model for the detection of oil spills using drone-captured RGB images of port environments.MethodsHyEff-UNet employs a modified version of the U-Net model with an EfficientNet-B4 model as an encoder is used as the model architecture to classify the pixels of the input image as oil, water, or other surface regions. A combined categorical cross-entropy and Dice loss function is used to jointly optimize pixel-wise classification and region overlap.ResultsThe results show that the complete architecture achieves the highest performance among the internally evaluated models, while the sequential ablation indicates that the largest performance improvement is observed after the hybrid categorical cross-entropy and Dice loss is introduced following the EfficientNet-B4 and attention-based architecture.ConclusionThese findings demonstrate the effectiveness of the complete architecture within the evaluated configurations. However, the sequential ablation study does not isolate the independent contribution of the hybrid loss. Furthermore, the measured inference latency indicates that further optimization and unmanned aerial vehicle (UAV)-specific benchmarking are required before real-time onboard deployment can be considered.
Related
- A Hybrid Framework for Air Pollution Forecasting using DSES, DLA, Self-Attention Transformer, and BiGRU with Bayesian Optimization
- Large-scale bioacoustic detection using semantic segmentation: a deep learning framework applied to fin whale calls in ocean-bottom seismometer recordings
- Image-Based Asymmetric Feature Fusion Using EfficientNet-B0 and MobileNetV3Large for Multi-Class Fish Disease Classification
- First-arrival time detection based on data-driven enhanced Akaike information criterion
- Integrated experimental and machine learning-driven multi-objective optimization for phenol removal via silica aerogel/ZSM-5 nanocomposite: Isotherm and kinetic analysis, thermodynamic insights and pareto-based process design
BibTeX
@article{dhanya2026hyeffunet,
title = {{HyEff-UNet: an attention-driven Efficientnet U-Net with hybrid loss optimization for UAV-based marine oil spill detection and segmentation}},
author = {Dhanya S. and Siva Priya M. S. and Aarthi B. and Prem Sankar N.},
journal = {Frontiers in Water},
year = {2026},
month = nov,
doi = {10.3389/frwa.2026.1932830},
url = {https://doi.org/10.3389/frwa.2026.1932830},
note = {CC BY},
}
RIS
TY - JOUR
TI - HyEff-UNet: an attention-driven Efficientnet U-Net with hybrid loss optimization for UAV-based marine oil spill detection and segmentation
AU - Dhanya S.
AU - Siva Priya M. S.
AU - Aarthi B.
AU - Prem Sankar N.
PY - 2026
DA - 2026/11/03
JO - Frontiers in Water
DO - 10.3389/frwa.2026.1932830
UR - https://doi.org/10.3389/frwa.2026.1932830
AB - IntroductionMonitoring oil spills in the port area is very difficult and can spread rapidly, resulting in considerable damage to the environment. This study proposes HyEff-UNet, a deep learning-based semantic segmentation model for the detection of oil spills using drone-captured RGB images of port environments.MethodsHyEff-UNet employs a modified version of the U-Net model with an EfficientNet-B4 model as an encoder is used as the model architecture to classify the pixels of the input image as oil, water, or other surface regions. A combined categorical cross-entropy and Dice loss function is used to jointly optimize pixel-wise classification and region overlap.ResultsThe results show that the complete architecture achieves the highest performance among the internally evaluated models, while the sequential ablation indicates that the largest performance improvement is observed after the hybrid categorical cross-entropy and Dice loss is introduced following the EfficientNet-B4 and attention-based architecture.ConclusionThese findings demonstrate the effectiveness of the complete architecture within the evaluated configurations. However, the sequential ablation study does not isolate the independent contribution of the hybrid loss. Furthermore, the measured inference latency indicates that further optimization and unmanned aerial vehicle (UAV)-specific benchmarking are required before real-time onboard deployment can be considered.
SN - 2624-9375
ER -