🌥️ Cloud Classification using Deep Learning

A deep learning project for classifying cloud types from images using Convolutional Neural Networks (CNN) and Transfer Learning.

📋 Table of Contents

• Overview
• Features
• Installation
• Quick Start
• Usage
• Project Structure
• Model Architectures
• Results
• Contributing
• License

🎯 Overview

This project implements a cloud classification system that can identify 7 different types of clouds:

• Cirriform clouds (Cirrus)
• Clear sky
• Cumulonimbus clouds
• Cumulus clouds
• High cumuliform clouds
• Stratiform clouds
• Stratocumulus clouds

The system uses state-of-the-art deep learning models including ResNet50, EfficientNet, and MobileNetV2 with transfer learning for accurate classification.

✨ Features

• 🔄 Multiple Model Architectures: Support for Simple CNN, ResNet50, EfficientNet, and MobileNetV2
• 📊 Comprehensive Evaluation: Confusion matrix, classification reports, and visualization tools
• 🎨 Data Augmentation: Built-in augmentation to improve model generalization
• 📈 Training Monitoring: Real-time metrics tracking with early stopping and learning rate scheduling
• 🔍 Easy Inference: Simple API for single image and batch predictions
• 📉 Visualization Tools: Automatic generation of training curves and prediction visualizations

🚀 Installation

Prerequisites

• Python 3.7 or higher
• TensorFlow 2.10 or higher
• CUDA (optional, for GPU acceleration)

Setup

1. Clone the repository:

git clone https://github.com/yourusername/cloud-classification.git
cd cloud-classification

2. Install dependencies:

pip install -r requirements.txt

3. Verify installation:

python test_setup.py

🏃 Quick Start

1. Explore the Dataset

python data_loader.py

This will display dataset information and visualize sample images.

2. Train a Model

# Train with ResNet50 (recommended)
python train.py --model resnet50 --epochs 50 --batch_size 32

3. Evaluate the Model

python evaluate.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --visualize

4. Make Predictions

python predict.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --image path/to/image.jpg

💻 Usage

Training

Basic Training

python train.py --model resnet50 --epochs 50

Advanced Training Options

python train.py \
    --model resnet50 \
    --epochs 50 \
    --batch_size 32 \
    --lr 0.001 \
    --save_dir models

Available Models

• simple: Simple CNN from scratch
• resnet50: ResNet50 with transfer learning (recommended)
• efficientnet: EfficientNetB0 (high accuracy)
• mobilenet: MobileNetV2 (lightweight, fast)

Training Options

• --no_aug: Disable data augmentation
• --unfreeze: Unfreeze base model for fine-tuning
• --batch_size: Batch size (default: 32)
• --lr: Learning rate (default: 0.001)

Evaluation

# Basic evaluation
python evaluate.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5

# With visualizations
python evaluate.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --visualize

Prediction

# Single image
python predict.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --image image.jpg

# Batch prediction
python predict.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --dir images/

# Top-K predictions
python predict.py --model models/resnet50_YYYYMMDD_HHMMSS/best_model.h5 --image image.jpg --top_k 5

Data Visualization

# Analyze dataset
python visualize_data.py --all

# Specific analyses
python visualize_data.py --stats    # Dataset statistics
python visualize_data.py --samples  # Visualize samples
python visualize_data.py --sizes    # Image size analysis

📁 Project Structure

cloud-classification/
├── clouds_train/              # Training dataset
│   ├── cirriform clouds/
│   ├── clear sky/
│   ├── cumulonimbus clouds/
│   ├── cumulus clouds/
│   ├── high cumuliform clouds/
│   ├── stratiform clouds/
│   └── stratocumulus clouds/
├── clouds_test/               # Test dataset
│   └── [same structure as train]
├── models/                    # Trained models (generated)
│   └── resnet50_YYYYMMDD_HHMMSS/
│       ├── best_model.h5
│       ├── final_model.h5
│       ├── class_mapping.json
│       ├── training_log.csv
│       └── training_curves.png
├── evaluation_results/       # Evaluation outputs (generated)
├── config.py                 # Configuration file
├── data_loader.py             # Data loading and preprocessing
├── model.py                   # Model architectures
├── train.py                   # Training script
├── evaluate.py                # Evaluation script
├── predict.py                 # Prediction script
├── visualize_data.py          # Data visualization tools
├── example_usage.py           # Usage examples
├── quick_start.py             # Quick start guide
├── test_setup.py              # Setup verification
├── requirements.txt           # Dependencies
└── README.md                  # This file

🏗️ Model Architectures

1. Simple CNN

A custom CNN architecture built from scratch with:

• 4 convolutional blocks
• Batch normalization and dropout
• Global average pooling
• Dense layers for classification

2. ResNet50 (Recommended)

Transfer learning with ImageNet pretrained ResNet50:

• Freeze base model option
• Fine-tuning support
• High accuracy

3. EfficientNetB0

EfficientNet architecture optimized for accuracy and efficiency:

• Compound scaling method
• State-of-the-art performance

4. MobileNetV2

Lightweight model suitable for mobile and edge devices:

• Depthwise separable convolutions
• Fast inference time

📊 Results

After training, you'll get:

Model Files

• best_model.h5: Best model based on validation accuracy
• final_model.h5: Final model after all epochs
• class_mapping.json: Class to index mapping
• training_log.csv: Training history
• training_curves.png: Visualization of training metrics

Evaluation Outputs

• confusion_matrix.png: Confusion matrix visualization
• class_metrics.png: Per-class precision, recall, F1-score
• prediction_samples.png: Sample predictions with images
• evaluation_report.txt: Detailed classification report

🔧 Configuration

Edit config.py to customize:

# Image settings
IMG_SIZE = (224, 224)
BATCH_SIZE = 32
EPOCHS = 50
LEARNING_RATE = 0.001

# Model settings
MODEL_NAME = 'resnet50'
PRETRAINED = True
FREEZE_BASE = True

# Data augmentation
USE_AUGMENTATION = True
AUGMENTATION_CONFIG = {
    'rotation_range': 20,
    'width_shift_range': 0.1,
    'height_shift_range': 0.1,
    'shear_range': 0.1,
    'zoom_range': 0.1,
    'horizontal_flip': True,
    'brightness_range': [0.8, 1.2]
}

📈 Tips for Better Results

1. Use Data Augmentation: Always enable augmentation to increase data diversity
2. Transfer Learning: Use pretrained models (ResNet50, EfficientNet) instead of training from scratch
3. Fine-tuning: After training with frozen base, unfreeze and fine-tune with lower learning rate
4. Hyperparameter Tuning: Experiment with different learning rates and batch sizes
5. Ensemble Methods: Combine multiple models for better accuracy

🐛 Troubleshooting

GPU Not Detected

import tensorflow as tf
print(tf.config.list_physical_devices('GPU'))

Out of Memory

• Reduce batch_size (e.g., --batch_size 16)
• Use a lighter model (MobileNet)
• Reduce image size in config.py

Training Too Slow

• Use GPU acceleration
• Increase batch_size if you have enough RAM
• Use data generators (already implemented)

Import Errors

pip install -r requirements.txt

📚 Examples

See example_usage.py for comprehensive usage examples:

python example_usage.py

This interactive script guides you through:

1. Data exploration
2. Data loading
3. Model creation
4. Training
5. Evaluation
6. Prediction

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• TensorFlow/Keras: Deep learning framework
• ImageNet: Pretrained model weights
• Dataset: Cloud images classification dataset

👤 Author

Thipv0302

• GitHub: @Thipv0302

📧 Contact

For questions or suggestions, please open an issue on GitHub.

---

Made with ❤️ for cloud classification

⭐ If you find this project useful, please consider giving it a star!