Imagine visiting a serene cove at low tide. To one side, a woman sings one of the beautiful songs of New Zealand’s Maori people. You look up to take in a towering, ornate totem — the work of dozens of skilled human hands — the Pou Kapua.
Forget HAL. When most of us think of AI, our minds turn to sci-fi movies depicting the work of machines. But after a stunning demo this week showcasing the work of traditional artists, maybe more people will see AI as a way to help share — and create — what makes us most human: art.
At the Game Developers Conference this week in San Francisco, we announced the release of new AI-powered materials and textures tools as part of GameWorks. It’s the latest in our effort to find new ways to bring new deep learning findings to film, professional and computer game graphics.
These tools allow developers and artists to use AI to automatically expand the variation of textures in their game; improve the image quality in games; and automatically generate textures and matching materials properties for real-world objects from photographs.
When we showed these off, the highlight was a stunning demo that makes clear how machine learning can help accelerate the field of photogrammetry, in which DSLR photos are combined to create richly detailed 3D models of real-world objects.
It’s a case study of how GPUs — and the CUDA parallel computing platform and programming model — are powering machine learning applications for artists and engineers working across industries to create amazing experiences.
A Global Effort
The story starts with a Slovak company, CapturingReality, which creates the highly optimized “RealityCapture Engine” that uses the power of CUDA to give artists photogrammetry results in minutes rather than hours. Such results would have been considered science fiction only a few years ago.
Using photogrammetry, they created a rich, sensory VR experience that allows users all around the world to experience a sacred site of the Maori people, the South Pacific island nation’s first inhabitants.
A Towering VR Experience
Measuring 80 feet in height and weighing more than 20 tons, the centerpiece of the site — Pou Kapua — is carved from 2,000-year-old kauri tree that’s native to New Zealand’s North Island.
The carving took the skill of over 35 master carvers and indigenous artists. It depicts numerous guardians, including Tane (Forest), Tawhirimatea (Winds), Tumatauenga (War), Ruaumoko (Volcanoes), and Rongo (Peace) and is topped with the demi-god Maui.
To capture the vibrant colors and meticulously carved detail in the totem, the firms took more than 15,000 photos. These were then processed to produce a highly realistic 3D model that perfectly matches the real totem now standing in Manukau, New Zealand.
It’s an example of how machine learning can combine with human creativity to provide an amazing new kind of experience. And the kind of experience that may soon change the way we all think about AI.