New in-development facial capture solution works without markers or specialised cameras
Nvidia has unveiled a new “photorealistic” facial capture demo showing off its FaceWorks technology.
Generated in real-time using commercially available GPUs and FaceWorks itself, the impressive video showed off two close-to-realistic digital head models of a man and a woman.
The technology is part of a scheme created in collaboration with the Institute for Creative Technology at the University of Southern California, in which developers have been building new systems to capture data to “within a tenth of a millimetre” without the need for markers of specialised cameras.
A special ‘light stage' tech uses photographic techniques to record the 3D shape of an actor’s face while also capturing various other elements to help properly represent human skin. This includes light reflections, reflectivity of oils in the skin and facial lines and bumps in the skin.
Nvidia claimed that complex skin shaders executing over 8,000 instructions per pixel to allow for realistic rendering, at full HD resolution, would create 82 billion floating point operations per second per frame, which would exceed 32gb for a few expressions.
The tech giant claimed that its FaceWorks tech however had reduced this to 300mb.
To view the FaceWorks tech demo in action, you can view the video below.
<iframe width="430" height="242" src="http://www.youtube.com/embed/5d1ZOYU4gpo" frameborder="0" allowfullscreen></iframe>