August 15, 2025Open Access

Rapid Flood Mapping from Aerial Imagery Using Fine-Tuned SAM and ResNet-Backboned U-Net

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Analysis reveals fine-tuned segment anything model improves flood mapping accuracy in aerial imagery, suggesting a new tool for damage assessment.

Fine-tuned Segment Anything Model achieved 96% accuracy with point prompts, enhancing flood mapping effectiveness.
Bounding box prompts showed lower accuracy at 82%, highlighting the need for precise boundary detection in flood imagery.
U-Net with ResNet-50 and ResNet-101 backbones showed accuracy improvements, reaching 88% with the deeper model.
Applications include rapid damage assessment for emergency teams and potential training data for machine learning models.

Shokati et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Analysis reveals fine-tuned segment anything model improves flood mapping accuracy in aerial imagery, suggesting a new tool for damage assessment.