-
Update README.md
-
Update README.md
-
Upload Aeris_Model.pth
# AERIS: Cloud Detection in Satellite Imagery
Automated cloud detection system for Landsat 8 satellite imagery using deep learning with uncertainty quantification.
## Overview
AERIS (Automated Environmental Remote Imaging System) detects clouds in 4-channel satellite images using a U-Net architecture with ResNet34 encoder. The model includes Monte Carlo Dropout for uncertainty estimation, making predictions reliable and trustworthy.
## Model Architecture
- **Base:** U-Net with ResNet34 encoder
- **Input:** 4-channel images (256×256) - Red, Green, Blue, Near-Infrared
- **Output:** Binary cloud segmentation mask
- **Uncertainty:** MC Dropout with 30 iterations
- **Framework:** PyTorch + segmentation-models-pytorch
## Performance Metrics
| Metric | Score |
|--------|-------|
| Validation IoU | 92.20% |
| Dice Coefficient | 94.28% |
| Precision | 92.15% |
| Recall | 96.73% |
| F1 Score | 94.28% |
| ECE (Calibration) | 0.70% |
## Training Details
- **Dataset:** 38-Cloud (16,800 trainin… -
initial commit
