It is unquestionable that photorealistic digital humans will become ubiquitous in society, whether in the form of AI assistants or as fictional characters on viral media or as our own virtual self for social interactions. While currently creating a convincing digital human involves an expensive and lengthy procedure from a team of VFX experts, in the near future, anyone will be able to create photorealistic human content at their fingertips. In this talk, I present techniques to create photorealistic digital humans using 3D computer graphics and deep learning. Using the Light Stage high-fidelity capture systems, I describe how we can achieve realistic rendering of an animated face in real-time that is accurate to the level of microns. By combining cutting edge 3D graphics and deep generative models, I present methods to model, animate, and render photorealistic 3D humans from minimal inputs to bring avatar digitization to everyone. I will also showcase the applications of deep generative models for lifelike digital humans for VFX, gaming and an autonomous virtual agent. While these are key technologies for the creation of consumer accessible virtual beings, they can also be misused for malicious purposes such as the spread of disinformation. To that end, I will discuss a method to detect advanced media forgeries such as deepfakes and our efforts to fight against them.
Speaker: Koki Nagano, NVidia Research
Register at weblink to receive connection information
Contact:Website: Click to Visit
Save this Event:iCalendar
Windows Live Calendar