Pneumonia Detection — Flask Web App (Chest X‑Ray CNN)

A simple Flask web application that performs pneumonia detection from chest X‑ray images using a CNN model. This README provides setup, run, and troubleshooting steps formatted for GitHub.

Quick links

• Project root: app.py, util.py, templates/, static/, models/oldModel.h5
• Dashboard: /dashboard (requires authentication)
• Prediction endpoint: POST /predict (expects base64 image payload)

1 — System requirements

• OS: Windows 10/11, macOS, or Ubuntu
• Python: 3.10 or 3.11 recommended
• RAM: ≥ 8 GB (TensorFlow can be memory hungry)
• Disk: ≥ 2 GB free
• GPU: optional (NVIDIA + CUDA for faster TF inference)

2 — Clone the repository

git clone https://github.com/<your-org>/<your-repo>.git
cd <your-repo>

3 — Create an isolated Python environment Using conda (recommended):

conda create -n pneumonia-ai python=3.10 -y
conda activate pneumonia-ai

Or using venv:

python -m venv venv
# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

4 — Install dependencies

pip install --upgrade pip
pip install -r requirements.txt

5 — Configure environment variables Create a .env file in the project root (same folder as app.py) with the following values:

SECRET_KEY=super-secret-key
SUPABASE_URL=https://your-supabase-url.supabase.co
SUPABASE_ANON_KEY=your-supabase-anon-key
GEMINI_API_KEY=your-google-gemini-api-key
COOKIE_SECURE=0

Do not commit .env to public repos.

6 — Place your trained model Put your Keras/TensorFlow .h5 model at:

models/oldModel.h5

Note: model input size must match the code (the app uses target size 64×64 by default). Adjust app.py preprocessing if your model requires a different size.

7 — Run the Flask app (development)

python app.py

Expected output (example):

 * Running on http://127.0.0.1:5002/

Open: http://127.0.0.1:5002/

8 — Using the app

• Visit the home page and sign in (Supabase/Google flow).
• Go to Dashboard.
• Upload an X‑ray image (JPG/PNG). The UI sends a base64 image to /predict.
• The server returns JSON: e.g. { "result": "PNEUMONIA" }.

9 — Production & deployment snippets Gunicorn (example):

# Simple run
gunicorn -b 0.0.0.0:5000 app:app

# With gevent
gunicorn -k gevent -w 1 app:app -b 0.0.0.0:5000

Docker (simple Dockerfile example):

FROM python:3.10-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY . .
EXPOSE 5000
CMD ["gunicorn", "app:app", "--bind", "0.0.0.0:5000"]

10 — Troubleshooting & common fixes

• tensorflow not found
  • pip install tensorflow==2.12.0
• Error loading model
  • Confirm models/oldModel.h5 exists and matches Keras/TensorFlow versions
• ModuleNotFoundError: dotenv
  • pip install python-dotenv
• Static files not loading
  • Confirm the static/ and templates/ directories are in the application root
• Memory issues (TF)
  • Increase instance memory or use smaller model; consider loading model lazily or using an external inference service

11 — Optional: environment.yml (Conda one‑command setup) Save as environment.yml and share with your team:

name: pneumonia-ai
channels:
  - defaults
dependencies:
  - python=3.10.9
  - pip
  - pip:
      - flask==3.0.3
      - gevent==24.2.1
      - gunicorn==22.0.0
      - requests==2.32.3
      - python-dotenv==1.0.1
      - tensorflow==2.12.0
      - keras==2.12.0
      - numpy==1.23.5
      - h5py==3.8.0
      - pillow==10.3.0

12 — Notes

• The app may lazy-load the model (recommended) to reduce startup memory usage — check app.py for model-loading logic.
• Never commit private API keys or secrets to a public repository.

Credits

• Original authors and contributors
• Designed by Raj Gajjar & Sadhu Dhwanika