Note: This video may require joining the NVIDIA Developer Program or login
GTC Silicon Valley-2019 ID:S9707:Volumetric Camera Systems and Light Fields
Tobias Chen(Volumetric Camera Systems)
We'll discuss the process of building a portable, cinematic-grade volumetric camera system powered by machine learning and describe a suite of scalable cloud processing tools using NVIDIA GPUs. We'll explain how a customized build of Unreal Engine makes it possible to play volumetric videos in VR in real time. We will also outline how volumetric videos can be streamed online through Sketchfab OSGB and Facebook GLTF format, established platforms with millions of users. In addition, we'll cover how we minimized the number of cameras needed for texturing by up-rezing our source videos from 4k to 8k using a GAN pre-trained neural network. This let us deliver cinematic-quality volumetric videos without long processing times or huge data requirements.