I’m an algorithm! Making ice cream everyday. When you favourite an image I learn from it to generate the next one. Learning day by day from your favourites! :icecream: Sunny day, 4 PM

I’m a three-minute-a-day artist. Every day I will go on Twitter and get the most favourited status from my account. From those I will create a new image. Learning day by day — a new image posted every day at 4 PM. Creating our perfect ice cream. I’m interested in knowing how an ice cream makes you feel. What are its dynamics? How do colours and shapes make it an ice cream? The program uses a parameterised circle, rectangle and triangle. It learns and evolves their placement, size, rotation and color by using a genetic algorithm. Using basic shapes help us study the relationship between the ice cream’s components: cone, scoop and chocolate flake. The first instalment is fully random. The program then uses previous ice creams to create a new one each day. It assumes that the more Twitter favourite an ice cream gets, the more its characteristics are important. It picks parents. The first parent is the ice cream with the most favourites. The second parent is randomly picked from all the ice creams that received favourites. The features of the parents are then mixed at random (genes crossover) to create the new ice cream DNA. Each shape has a random color picked from a subset of all possible colours. These subsets evolve over generations.

Read more on Medium

You will need:

• Go
• Python 2.7
• Redis

pip install -r requirements.txt
make
./bin/make/icecreamd

docker-compose build
docker-compose run botmachine icecreamd

Point your browser to: localhost:2003

You will see a gallery of ice cream. Reload the page to create a new one. Click once or multiple times on an image to favourite it.

After building the bin folder will cointain several tools:

• icecreamd: a web server serving the ice cream web app
• makeicecream: a tool to generate new ice cream or get an SVG from the database
• tweeticecream: a tool to tweet!
• syncicecream: a tool that will query Twitter for the favourite count and update the database for every ice cream.
• freezer2dot: a tool to generate a Graphviz visualisation of the database.
• freezer2json: a tool to dump the database as a JSON file.

In order to tweet one must fill the .env file with the following environment variables:

• TWITTER_CONSUMER_KEY
• TWITTER_CONSUMER_SECRET
• TWITTER_ACCESS_TOKEN
• TWITTER_ACCESS_TOKEN_SECRET

Copyright (c) 2016 François Hoehl

MIT License