Enhancing flood segmentation in aerial imagery using a dual-decoder efficient U-Net with CBAM and post-hoc explainable AI technique

View Full Paper

Ask AI

Bookmark

Authors

Discussion

This framework improves flood segmentation accuracy and interpretability in UAV imagery, suggesting better disaster response interventions.

DualEff-CBAMNet achieves 94.89% accuracy and 94.67% recall in flood segmentation, highlighting its effectiveness.
The model surpasses popular architectures like U-Net and DeepLabV3+, indicating superior performance in identifying flood areas.
Using a hybrid loss function and advanced augmentation techniques, the methodology addresses class imbalance and boundary detection.
LIME analysis demonstrates the model's interpretability by clarifying the influence of superpixels on predictions, revealing possible limitations.

Purohit et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

This framework improves flood segmentation accuracy and interpretability in UAV imagery, suggesting better disaster response interventions.