Presentation

<
GenUSD: 3D scene generation made easy
SessionReal-Time Live!
DescriptionWe present a text-to-USD method that creates 3D scenes with Universal Scene Description (USD). Given a text prompt, the method automatically generates a 3D scene graph and objects in USD format. The whole generation process takes less than 2 minutes, making the 3D scene generation easy.
Contributors
Tsung-Yi Lin
NVIDIA
Chen-Hsuan Lin
NVIDIA
Yin Cui
NVIDIA
Yunhao Ge
NVIDIA
Seungjun Nah
NVIDIA
Arun Mallya
NVIDIA
Zekun Hao
NVIDIA
Yifan Ding
NVIDIA
Hanzi Mao
NVIDIA
Zhaoshuo Li
NVIDIA
Yen-Chen Lin
NVIDIA
Xiaohui Zeng
NVIDIA
Qinsheng Zhang
NVIDIA
Donglai Xiang
NVIDIA
Qianli Ma
NVIDIA
J.P. Lewis
NVIDIA
Jingyi Jin
NVIDIA
Pooya Jannaty
NVIDIA
Ming-Yu Liu
NVIDIA
Fangyin Wei
NVIDIA
Jiashu Xu
NVIDIA
Event Type
Real-Time Live!
TimeTuesday, 30 July 20246:00pm - 7:45pm MDT
LocationBluebird Ballroom
Session TimeTuesday, 30 July 20246:00pm - 7:45pm MDT
LocationBluebird Ballroom
Interest Areas
New Technologies
Recordings
Livestreamed
Recorded
Keywords
AI
Machine Learning
Registration Categories
Full Conference
Full Conference Supporter
Virtual Access
Exhibitor Full Conference
Tuesday
Next PresentationNext PresentationMesh Mortal Combat: Real-Time Voxelized Soft-Body Destruction