A CNN model for the MIML dataset built with FastAI, ResNet-50 and Flask

In this repository you'll find:

• My Jupyter Notebook file for the training of a ResNet-50 based CNN using FastAI
• A ready to use REST-API in Flask, which returns predictions for any given image

Using the Jupyter-Notebook file

You need to have fastai>=1.0.0 installed.

Setup the REST-API

Please note, that i put my weights.pkl file in the rest-api folder. If you modified the model, you have to export your weights and replace the file.

Install requirements

pip install -r requirements.txt

Start server

FLASK_APP=rest-api/server.py flask run

Make a demo request

python request.py --file example.jpg
# Returns: {'result': 'desert'}