LifeLex / pneumonia-detection

Web app for detecting pneumonia from x-ray images

Detecting Pneumonia from X-ray Scans

I am using the Chest X-Ray Images (Pneumonia) dataset from Kaggle to train a simple CNN for pneumonia detection. The dataset splits pneumonia into two categories: viral and bacterial. The model does not differentiate between these two categories.

Model Deployment

Classifier is deployed via FastAPI + Docker. The endpoint accepts uploaded files and infers from the image whether there is pneumonia.

Next steps: push to a public URL.

Setting up Locally

Steps to spin the server on your local network:

1. Setup Docker
2. Run pip install -r requirements.txt
3. Create a Kaggle account and generate an API token from kaggle.com/USERNAME/account (this will prompt you to download a kaggle.json file which contains the credentials)
4. a) Set the Kaggle credentials as environment variables with:

export KAGGLE_USERNAME = [kaggle username]
export KAGGLE_KEY = [generated key]

4. b) OR use direnv to populate credentials in .envrc (see .envrc.example for formatting)
5. Run python classifier/train.py to train the model (Note: this step is optional as repo already comes with model weights)
6. Build Docker container using docker build . -t app
7. Run container using docker run -d -p 8080:5000 app

This deploys the app to local network http://localhost:8080/. You may have to replace localhost with your ip address if it doesn't work.

Making Requests

The easiest way to make a POST request is to go to http://localhost:8080/docs and uploading an image from the /pneumonia/predict.