AVACA
Artificial Intelligence Game for Still Life Composition. AVACA is a made-up name for a game available for free on Google Play, which is about competing with Artificial Intelligence to find the best composition for a still life painting.
available here to download on Android devices.
Based on the work of the Italian artist Giorgio Morandi, this game challenges you to achieve the best composition of ordinary items for painting. You can move and rotate the objects using your fingers and you need to be fast to beat the artificial intelligence. The best score is calculated using advanced neural networks and artificial intelligence had learned to compose items for composition over long hours of training. Some visual tools are provided to help you find the best composition. Help the AI to learn more about artistic aesthetics in visual composition by playing, your contribution will help the AI in the next generation of training.
This game is related to the paper https://direct.mit.edu/leon/article-abstract/55/1/57/99850/A-Machine-Learning-Application-Based-on-Giorgio?redirectedFrom=fulltext Guido Salimbeni, Frederic Fol Leymarie, William Latham; A Machine Learning Application Based on Giorgio Morandi Still-Life Paintings to Assist Artists in the Choice of 3D Compositions. Leonardo 2022; 55 (1): 57–61. doi: https://doi.org/10.1162/leon_a_02073.
It is part of my research around the use of AI to help creative practice. In this specific case, to help to find a best composition for a traditional still life painting.
By leveraging the features of a game engine framework, the system can analyse both the rendered image on the screen and obtain information on the position, depth and volume of the objects in the three-dimensional digital environment. Finding the optimal arrangement of visual elements for a painting is a manual and time-consuming activity that the painter pursues by applying general principles of design and personal aesthetic intuition. The system uses a three-dimensional digital prototype of the objects to paint or render in a design and proposes a computational technique to assist the painter or designer in the selection of the most aesthetically pleasing composition for a painting job. In this article, we illustrate how our system uses an evolutionary algorithm (EA) and four artificial neural networks to automate and speed up the creative process of compositional choice. Finally, we compare the results of our system with those produced by a method already used in commercial applications.