Art Done Quick

Casual creation of abstract art

This project aims to build a casual creator app with which people can make abstract pieces of art really quickly in a fun way. Through harnessing the latest developments in creative AI and machine vision, the app focuses on having fun while making stuff, not in producing quality output to a professional standard. The app will be used study how people interact with casual creators and to push forward our understanding of computational creativity.

Art Done Quick uses generative technologies developed over more than a decade to produce a 500×500 pixel, high-quality, novel abstract art image in about 1/10th of a second, which is expanded to a 2000×2000 pixel version in less than a second. Images are generated by moving particles around the canvas, re-colouring them and then drawing shapes at each particle’s location and colour. Blurring is added along the way to add depth and sophistication to the images.

A video showing the workings of Art Done Quick

Users are able to choose images to mutate by double-tapping them to produce variations. They can also cross-over two images to produce children, by combining the mathematical functions of the two parents. The app allows users to post-process the images with image filters, text and stickers, as well as collaging operations – with all the options chosen with a fun-first design methodology. The alterations are passed on to the mutations and children.

The final part of the app is the machine vision system, which is trained to identify the content of photos taken in the real world – and categorizes the abstract images according to what it ‘sees’ them as. While these are false-positives in a machine learning sense, they have a lot of artistic and entertainment value in the context of casual creators, and can encourage users to look again at an image.

Along with other casual creators, Art Done Quick will be used in psychology experiments as part of a PhD project to evaluate the states of mind people get into with casual creators, in order to understand the overlap of how people play and how they create. The findings will also pay into best practice guidelines for the engineering of casual creators, building on the work of Kate Compton, who identified and popularised the category of software apps in her PhD. The app will ready for commercial release shortly.

Project members

Simon Colton

 

Outcomes

Shown at SensiLab Open House (Dec 2019). The installation, ‘Can You See What I Can See?’, includes The Painting Fool software, which controls the app in order to make art in a fully automated way.