CloudMorph Lens: Cloud Classification & Segmentation Platform

Overview

CloudMorph Lens is a robust, production-ready deep learning platform for automated cloud type classification and segmentation from satellite or aerial imagery. Leveraging state-of-the-art computer vision models (ResNet50 for classification, U-Net for segmentation), StratoVision provides both a Python API and a Flask-based web interface for rapid experimentation, integration, and deployment.

---

Features

• Multi-label Cloud Classification
  Accurately predicts the presence of multiple cloud types (cirrus, cumulus, stratus, nimbus, altostratus) in a single image.

• Cloud Segmentation
  Generates pixel-level masks to delineate cloudy vs. non-cloudy regions.

• Explainability
  Integrated Grad-CAM visualizations for model interpretability.

• RESTful API
  Exposes endpoints for classification, segmentation, and overlay generation.

• Web UI
  User-friendly interface for uploading images and visualizing results.

• Synthetic Data Pipeline
  Includes utilities for generating and visualizing synthetic MODIS-like cloud images for rapid prototyping.

---

Directory Structure

model-2/
│
├── app.py                  # Flask application (API + UI)
├── cloud_model.pth         # Trained classification model weights
├── cloud_seg_model.pth     # Trained segmentation model weights
├── m2_StratoHack_Cloud.ipynb # End-to-end Jupyter notebook (training, inference, API)
├── requirements.txt        # Python dependencies
├── snap-1.png              # Example image
├── templates/
│   └── index.html          # Jinja2 template for web UI
└── uploads/
    └── ...                 # Uploaded images (runtime)

---

Quick Start

1. Clone the Repository

git clone https://github.com/your-org/stratovision.git
cd stratovision/model-2

2. Install Dependencies

It is recommended to use a virtual environment:

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt

3. Run the Flask App

python app.py

Visit http://localhost:5000 in your browser.

---

API Endpoints

POST /predict_class

• Description: Predict cloud type probabilities for an uploaded image.
• Request:
  multipart/form-data with a file field (image).
• Response:
  JSON with class probabilities.

POST /predict_mask

• Description: Generate a binary cloud mask for an uploaded image.
• Request:
  multipart/form-data with a file field (image).
• Response:
  PNG image (cloud mask).

POST /cloud_map

• Description: Overlay predicted cloud mask on the original image.
• Request:
  multipart/form-data with a file field (image).
• Response:
  PNG image (overlay).

---

Model Details

• Classification:

  • Architecture: ResNet50 (multi-label head)
  • Output: Probability for each cloud type

• Segmentation:

  • Architecture: U-Net
  • Output: Binary mask (cloud vs. non-cloud)

• Explainability:

  • Grad-CAM for visualizing model attention

---

Training & Experimentation

All training, evaluation, and API code is available in m2_StratoHack_Cloud.ipynb.
The notebook is fully self-contained and can be run end-to-end for reproducibility.

---

Example Usage

Python (requests)

import requests

with open("snap-1.png", "rb") as f:
    files = {"file": f}
    response = requests.post("http://localhost:5000/predict_class", files=files)
    print(response.json())

cURL

curl -X POST -F "file=@snap-1.png" http://localhost:5000/predict_class

---

Contributing

Contributions are welcome! Please open issues or submit pull requests for improvements, bug fixes, or new features.

---

License

This project is licensed under the MIT License.

---

Contact

For questions, support, or collaboration, please contact atharvklhare18@gmail.com.

---