Accelerating 3D simulation workflows with NVIDIA Omniverse and DGX Cloud on Azure
summary
A complete pipeline demonstrates synthetic dataset generation, inspection, and model training using Omniverse Replicator for randomized parking‑lot scenes that output RGB plus ground truth like depth, velocity, instance masks and 2D/3D boxes, streamed from Omniverse Cloud and stored to Azure, then trained via DGX Cloud (H100) with Base Command and Weights & Biases. The demo programmatically varies vehicles and pedestrians, shows training loss falling over epochs, and produces correct semantic labels (cars magenta, pedestrians cyan).
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:31Industrial Enterprises are Racing to Digitalize
- 2:36NVIDIA Cloud Services
- 4:11NVIDIA Omniverse
- 4:53Introduction to Autonomous Vehicles
- 5:35AV Development: Massive Al and Data Challenge
- 7:54Simulation and Synthetic Data are Essential for AVS
- 9:04DRIVE Sim - Virtual Worlds for Autonomous Vehicles
- 10:17Omniverse Replicator
- 11:22Workflow on Omniverse Cloud - Data Generation
- 13:27Workflow on Omniverse Cloud - Data Inspection
- 14:03Workflow on DGX Cloud - Training
- 15:35Workflow on DGX Cloud - Testing
- 16:33Generative Al for Simulation and Synthetic Data
related talks
23mFrom Gaming to the Metaverse
16mImagine a new era of digitalization in manufacturing with NVIDIA
22mQuickly deploy NVIDIA optimized generative AI on Azure AI Studio
20mAccelerating AI pipelines w/ NVIDIA on AWS: An autonomous driving case
14mProcedural Simulation, AI creativity and stochastic modeling. Using Azure Cosmos DB to make games.
33mNext-generation tools for synthetic data and AI training
50mAccelerate development and deployment of Gen AI with NVIDIA on Azure
30mFast and Furious: Practice in Horizon Robotics on Large-scale End-to-end Model Training