Google has released a prototype of an AI app called Chimera Painter that turns your drawings into fantasy creatures.
The team decided to build a complete fictional digital paper game made up of a hundred creatures, where players collect different fictional creatures and make them fight.
The search giant developed the Chimera Painter app to produce a group of creatures for the card game, so that each of them could merge with each other to produce new creatures very quickly.
The tool uses machine learning technology to create images based on rough graphics for users.
This type of use became relatively common in machine learning, as Nvidia did it with landscapes, as did MIT and IBM with buildings.
Chimera Painter is powered by GANs, which generate new content by pitting two neuron networks against each other to generate new images, while determining which ones were artificially generated.
The tool’s developers trained generative GANs across 10,000 samples of computer-engineered creatures using 3D models provided by the Unreal game engine.
Each image was paired with a segmentation map that divides the imaginary creatures into anatomical parts, such as: claws, nose, legs, etc.
This method allowed generating new creatures by combining different parts of the anatomical parts of the imaginary creatures.
Google believes that the tool can be used to generate fictional creatures ready for video games.
The tool can be used to customize previously uploaded designs, import ready-made images, or generate new images through the tool.
The Chimera Painter tool is a digital artist’s paintbrush that reduces the amount of time required to create high-quality art without affecting artistic choice.
We hope the generative adversarial network (GAN) models and Chimera Painter tool will inspire others to think differently about technical streak, and what one can generate when using machine learning as a paintbrush, the app developers wrote in a blog post.
They added: The idea is to find a paintbrush that helps the artist more than just being a tool, and in the event that these programs and their ilk become popular, they may reduce the time needed to produce high-quality art in the future.