Flood Detection using U-Net with Attention Mechanism

This project implements a deep learning solution for flood detection using satellite imagery. The model uses a U-Net architecture enhanced with attention mechanisms to accurately segment and identify flooded areas in images.

Features

• U-Net architecture with attention gates for improved segmentation
• Data augmentation pipeline for robust model training
• Binary segmentation for flood detection
• Comprehensive evaluation metrics including IoU and Dice coefficient
• Support for custom dataset integration

Requirements

• Python 3.7+
• TensorFlow 2.x
• OpenCV
• NumPy
• scikit-learn
• Matplotlib

## Dataset Structure

The project expects the following dataset structure:

Dataset/
├── images/
│   ├── 1.jpg
│   ├── 2.jpg
│   └── ...
└── labels/
    ├── 1.png
    ├── 2.png
    └── ...

Model Architecture

The model implements a U-Net architecture with the following key components:

• Encoder path with multiple convolutional blocks
• Attention gates for feature refinement
• Decoder path with skip connections
• Binary segmentation output

Training

To train the model:

1. Place your dataset in the appropriate directories
2. Adjust the hyperparameters in the notebook if needed
3. Run the training cells in the Jupyter notebook

Evaluation

The model is evaluated using:

• IoU (Intersection over Union)
• Dice coefficient
• Accuracy metrics
• Confusion matrix

Results

The model achieves high accuracy in flood detection with:

• Precise segmentation of flooded areas
• Robust performance across different image conditions
• Efficient processing of satellite imagery