Tuberculosis Detection

This is an end-to-end tuberculosis classification tool that runs on Google Cloud.

It contains

• model training pipeline using a DenseNet121 architecture was built. Then the model is dockerized to be put on GCP.
• simple front-end where chest X-rays are given as input
• back-end
  • a cloud run to call the model with the input image
  • the dockerized model which is used by the cloud run to get the predictions

Demo

demo3.mov

Installation

Prerequisite: pyenv

https://github.com/pyenv/pyenv-installer

On macOS you can use brew, but you may need to grab the --HEAD version for the latest:

brew install pyenv --HEAD

or

curl https://pyenv.run | bash

And then you should check the local .python-version file or .envrc and install the correct version which will be the basis for the local virtual environment. If the .python-version exists you can run:

pyenv install

This will show a message like this if you already have the right version, and you can just respond with N (No) to cancel the re-install:

pyenv: ~/.pyenv/versions/3.8.6 already exists
continue with installation? (y/N) N

Prerequisite: direnv

https://direnv.net/docs/installation.html

curl -sfL https://direnv.net/install.sh | bash

Developer Setup

If you are a new developer to this package and need to develop, test, or build -- please run the following to create a developer-ready local Virtual Environment:

direnv allow
python --version
pip install --upgrade pip
pip install poetry
poetry install

Remarks

In a more professional settings, this repo should be split into multiple pieces. A good split may be

• model development repo
• inference + dockerization + deployment repo
• frontend

Because model development dependencies will make the inference pipeline unnecessarily larger. Similarly, frontend dependencies (currently flask and flask-cors) make the docker image larger.

To prevent this issue, I made all the model development libraries as dev dependencies. So that they are not included in the docker image. But this is just a trick, not the preferred way.

Data and Credit

The data for this model training is obtained from Kaggle:

https://www.kaggle.com/datasets/tawsifurrahman/tuberculosis-tb-chest-xray-dataset

Kudos to the team and their amazing work for collecting such a useful dataset!