🫁 Chest Cancer Classification App

A full-stack deep learning project that classifies chest X-ray images into various cancer categories using TensorFlow/Keras. It integrates data versioning (DVC), model tracking, Docker, and a Flask-based web interface, with CI/CD workflows through GitHub Actions and remote storage on DAGsHub.

---

📌 Features

• 🧠 Deep Learning CNN model using transfer learning (EfficientNet)
• 📦 Modular codebase with pipeline stages (data_ingestion, base_model, training, evaluation)
• 🚀 Real-time prediction web app using Flask
• 📈 Model evaluation and confidence scores
• 🔁 Reproducible experiments with DVC
• 🐳 Dockerized for seamless deployment
• ☁️ Remote artifact storage using DAGsHub
• ⚙️ GitHub Actions for CI/CD automation

---

📁 Project Structure

Click to expand/collapse

└── arpitkadam-chest-cancer-classification-app/
    ├── README.md
    ├── __init__.py
    ├── app.py
    ├── demo.py
    ├── Dockerfile
    ├── dvc.lock
    ├── dvc.yaml
    ├── LICENSE
    ├── main.py
    ├── params.yaml
    ├── requirements.txt
    ├── setup.py
    ├── template.py
    ├── .dockerignore
    ├── .dvcignore
    ├── artifacts/
    │   ├── base_model/
    │   │   ├── base_model.h5
    │   │   └── updated_base_model.h5
    │   ├── data_ingestion/
    │   │   └── data.zip
    │   ├── model_evaluation/
    │   │   └── evaluation_scores.json
    │   └── model_trainer/
    │       ├── trained_model.h5
    │       └── model_history/
    │           └── model_history.json
    ├── config/
    │   ├── config.yaml
    │   ├── model.yaml
    │   └── schema.yaml
    ├── Research/
    │   └── research.ipynb
    ├── src/
    │   ├── __init__.py
    │   ├── components/
    │   │   ├── __init__.py
    │   │   ├── base_model.py
    │   │   ├── data_ingestion.py
    │   │   ├── model_evaluation.py
    │   │   └── model_trainer.py
    │   ├── configuration/
    │   │   ├── __init__.py
    │   │   └── configuration.py
    │   ├── constants/
    │   │   └── __init__.py
    │   ├── entity/
    │   │   ├── __init__.py
    │   │   ├── artifact_entity.py
    │   │   └── config_entity.py
    │   ├── exception/
    │   │   └── __init__.py
    │   ├── logger/
    │   │   └── __init__.py
    │   ├── pipeline/
    │   │   ├── __init__.py
    │   │   ├── prediction_pipeline.py
    │   │   └── training_pipeline.py
    │   └── utils/
    │       └── __init__.py
    ├── templates/
    │   ├── home.html
    │   ├── importance.html
    │   └── prediction.html
    ├── .dvc/
    │   └── config
    └── .github/
        └── workflows/
            └── main.yaml

🚀 Getting Started

1. Clone the Repository

git clone https://github.com/ArpitKadam/Chest-Cancer-Classification-App.git
cd Chest-Cancer-Classification-App

2. Create and Activate Virtual Environment

conda create -n tensor python=3.10 -y
conda activate tensor

3. Install Dependencies

pip install -r requirements.txt

4. Run Training Pipeline

python main.py

5. Launch Web Application

python app.py

Navigate to: http://127.0.0.1:8080

---

📦 Docker Usage

Build Docker Image

docker build -t arpitkadam/chest-cancer-app:latest .

Run Container

docker run -p 8080:8080 arpitkadam/chest-cancer-app:latest

---

💾 DVC Workflow

Setup Remote (only once)

dvc remote add -d storage https://dagshub.com/ArpitKadam/Chest-Cancer-Classification-App.dvc

Reproduce pipeline and push data

dvc repro
dvc push

---

⚙️ CI/CD

GitHub Actions configured in .github/workflows/main.yaml for:

• Linting and testing
• Pushing data to DAGsHub
• Auto-retraining (optional)

---

🧪 Model Metrics

Evaluation scores are saved in: artifacts/model_evaluation/evaluation_scores.json

Click to expand/collapse

{
    "loss": 0.5021160244941711,
    "accuracy": 0.7684127163887024
}

---

📂 DVC Tracked Artifacts

• data_ingestion/data.zip
• base_model/base_model.h5
• model_trainer/trained_model.h5
• model_history/model_history.json
• model_evaluation/evaluation_scores.json

---

📄 License

This project is licensed under the Apache-2.0 license.

---

✨ Author

Arpit Sachin Kadam

• 💼 LinkedIn
• 🌐 Portfolio
• 📦 GitHub
• 🧪 DAGsHub

---

🙌 Contributions

Contributions are welcome! Please open issues or submit a PR.

---

🔧 Tech Stack

• Deep Learning: TensorFlow/Keras, VGG19
• Web Framework: Flask
• Data Versioning: DVC
• MLOps: DAGsHub
• Containerization: Docker
• CI/CD: GitHub Actions
• Frontend: HTML, CSS, JavaScript

---

📊 Usage Examples

Command Line Prediction

python main.py

Web Interface

1. Upload chest X-ray image
2. Get classification results with confidence scores
3. View model interpretation and importance maps

---

🚨 Important Notes

• Ensure you have sufficient GPU memory for training
• The model is trained on chest X-ray images - ensure input images are in the correct format
• For production deployment, consider implementing additional security measures
• Regular model retraining is recommended as new data becomes available

---

📈 Future Enhancements

• Add more cancer types for classification
• Implement model explainability features (LIME/SHAP)
• Add user authentication and session management
• Integrate with medical imaging standards (DICOM)
• Add batch processing capabilities