This is a little portfolio - construction ongoing.

A simple image classifier, which spits out a result if it's more than 90% confident, otherwise professes ignorance (cf. 'Artifical Stupidity', Vincent Warmerdam at PyData conference 2019: we shouldn't be extrapolating about things far outside the original training data, even though models let you do so.

Hosted here

Jupyter notebook where I trained the model.

Git clone this repo, and cd into it.

Create and activate your favourite virtual environment for Python.

At least on Windows, pytorch must be installed before fastai. Choose the correct installation here

Then:

pip install -r requirements.txt

Set the environment variables by creating .env at the root directory. This project doesn't need high security so I've just pasted the needed environment variable here to allow anyone to use the project.

python manage.py migrate
python manage.py collectstatic
python manage.py runserver