Presentation

Dive Into 3D Deep Learning, Physics Simulation, and Interactive 3D Prototyping With NVIDIA Kaolin Library
DescriptionNVIDIA Kaolin Library is a PyTorch library for 3D deep learning, offering a continuously expanding collection of tools that make 3D research not only faster, but more interactive. Using hands-on code samples, this lab will introduce the new physics simulation module, based on the Simplicits method (published this SIGGRAPH), which allows running physics simulation over any 3D representation. Specifically, the lab will show how to set up a physics simulation of a 3D Gaussian Splat reconstruction, interact with it, and visualize it in a jupyter notebook in just a few lines of code. In addition, we'll walk the audience through code samples covering key features of Kaolin, such as differentiable rendering with PBR materials, consistent geometry imports into PyTorch, debugging and inspection of custom 3D render functions, and much more. Finally, we'll discuss how to use Kaolin Library API to connect 3D AI research to the live USD scene inside NVIDIA Omniverse, showing a released code sample for Diffusion Texture Painting (also published this SIGGRAPH).
Event Type
Industry Session
TimeThursday, 1 August 20242:15pm - 3:15pm MDT
LocationRoom 709
Keywords
NVIDIA Lab
Registration Categories
Exhibitor
Full Conference
Full Conference Supporter
Experience
Exhibits Only
Exhibitor Full Conference
Exhibitor Experience
Thursday