genie.io

Local Development

Pre-requisites:

• Set KAGGLE_USERNAME and KAGGLE_KEY environment variables with your Kaggle credentials
• Set FOOD_API_ID, FOOD_API_SECRET, RECIPES_API_ID and RECIPES_API_SECRET environment variables with your Edamam API credentials
• Download food-101 images datasets from kaggle using mount_food_101.py script

Get started with genie running following commands:

• Create a virtualenv:

python -m venv venv && source venv/bin/activate

• install requirements:

pip install -r requirements.txt

Note: app requires python version>=3.11 (according to vermin)

• run genie in CLI mode:

python genie_driver.py --cli-mode

• run genie in API mode:

python genie_driver.py --api-mode

• train a new genie CNN model with specific number of food types (default=101):

python genie_driver.py --new-genie --num-types=<number of supported food types>

• train a new genie CNN model with GPU adaptation:

python genie_driver.py --new-genie --gpu

Container Image

Build and run application docker image

• Create secrets.env file with Kaggle and Edamam API credentials
• Build a new image tag locally:

docker build -t genie-io:<image_tag> .

• Run app in a container (API mode by default):

docker run -it -p 5000:5000  genie-io:<image_tag>