Project Material World

Using AI to convert pictures into digital materials

Traditionally, digital materials can either be designed from scratch using procedural tools like Substance Designer, or reconstructed from multiple images using computer vision methods. While extremely powerful, both methods are difficult to set up and require expert domain knowledge.

Project Material World simplifies the process of generating digital materials by computing complete, high-resolution digital material models from images — even photos taken with phones. This significantly simplifies the creation of 3D assets for which digital material models are key to achieving realism.

Material World uses a neural architecture trained on procedurally augmented data to provide an extremely simple material acquisition process. Starting from a single picture taken under natural lighting conditions, Material World avoids the need for a complex capture setup or extensive user inputs. Material World can even batch process photos in order to increase the productivity of 3D artists even more.

You can digitize the outside world with Image to Material filter in Substance Alchemist today. Take your photographic data and turn it into a realistic, usable material. Powered by AI, the Image to Material filter generates a perfect albedo, as well as normal and displacement maps which closely match the original real-life surface.

UpdatedOct 20, 2020
CategoryMachine Learning3D

3DI Research