As a student at Queensland University of Technology (QUT) in Australia, Emily Bohmer was torn between pursuing the creative arts or sciences.
She then discovered industrial design, which allowed her to delve into research and programming while exploring visualization workflows such as drawing, animation, and 3D modeling.
Now, Böhmer is putting her skills into practice as a design intern at the BMW Group Technology Office in Munich. The team uses NVIDIA Omniverse, a platform for developing and connecting 3D tools and applications, and Universal Scene description – also known as OpenUSD – to enhance their synthetic data generation pipelines.
Boehmer creates photorealistic 3D assets that can be used with SORDI.ai, which is short for Synthetic Object Recognition Dataset for Industries. Published by BMW Group, Microsoft and NVIDIA, SORDI.ai helps developers and researchers simplify and accelerate training AI for production. To automate the image generation process, the team developed an extension based on Omniverse Replicator, a software development kit for creating custom synthetic data generation tools.
As part of the SORDI.ai team, Boehmer uses Blender and Adobe Substance Painter to design 3D assets with high levels of physical and photorealistic fidelity, helping to ensure synthetic data can be used to train AI models efficiently.
All of the assets Boehmer creates are used to test and simulate autonomous robots on the NVIDIA Isaac Sim platform, which provides developers with a set of synthetic data generation capabilities that can run realistic and physically accurate virtual environments.
Create realistic 3D assets to train AI
As a design intern, Boehmer’s main duties are 3D animation and modeling. The process begins by taking photos of the target objects. Then, you use the 2D images as references by aligning them with the 3D models in Blender.
3D objects can be composed of thousands of polygons, so Boehmer creates two copies of the original — one with a low number of polygons and one with a higher number of polygons. Details of the high-poly version can be incorporated into the low-poly model, which helps preserve more detail so the original looks realistic.
Once the 3D assets are created, Boehmer uses models to start assembling the scenes. Her favorite aspect of the Omniverse platform is the flexibility of the USD, because it allows her to make changes to the 3D models easily.
USD workflows have enabled the BMW Group’s design teams to create many different scenes using the same components, as they can easily access all USD files stored on the Omniverse Nucleus. When creating parts of the scene, Boehmer pulls from dozens of USD models from SORDI.ai and adds them to scenes that other designers will use to assemble larger factory scenes.
Boehmer only has to update the original asset’s USD file to automatically apply the changes to all reference files it contains.
“It’s great to see USD support for both Blender and Substance Painter,” she said. “When I create 3D assets using USD, I can be confident that they will look and behave as expected in the scenes they will be placed in.”
Factory construction scenes using synthetic data
The Isaac Sim platform is an essential part of the SORDI.ai team’s workflow. It is used to develop pipelines that use generative AI and procedural algorithms to create a 3D scene. The team also developed an extension based on Omniverse Replicator that automates in-scene randomization when generating synthetic images.
“The role of design interns like me is to realistically design and install the assets used in the scenes built in Isaac Sim,” Bohmer said. “The more realistic the assets are, the more realistic the synthetic images are and the more effective they are in training AI models for real scenarios.”
Data annotation — the process of labeling data such as images, text, audio, or video with relevant tags — makes it easier for AI to understand the data, but the manual process can be incredibly time-consuming, especially for large amounts of content. SORDI.ai addresses these challenges by using synthetic data to train AI.
When you import assets into Omniverse and create USD versions of files, Boehmer tags them with the appropriate data label. Once these assets are grouped together in a scene, they can use Omniverse Replicator to create images that are automatically annotated with the original labels.
With SORDI.ai, designers can set up scenes and create thousands of annotated images with just one click.
Boehmer will be a guest on the Omniverse livestream on Wednesday, September 20, where she will demonstrate how to use Blender and Substance Painter in Omniverse for synthetic image generation pipelines.
Join creation
Anyone can create their own Omniverse plugin or connector to enhance their 3D workflow and tools. Creators and developers can download Omniverse for free, and enterprise teams can use the platform for their 3D projects.
Check out artwork from other “carnivores” and submit projects in the gallery. Learn how content creators are using OpenUSD to accelerate a variety of 3D workflows in the latest release of OpenUSD All Stars. And bring your workflow to the Omniverse using software from Adobe, Autodesk, Blender, Epic Games, Rellusion, and more.
Get started with NVIDIA Omniverse by downloading the standard license freeOr learn how Comprehensive enterprise Your team can be linked. Developers can Get started with Omniverse Resources and learning about them OpenUSD. Discover the The growing ecosystem of 3D tools Connected to the Omniverse.
Stay up to date on the platform by subscribing to the newsand follow NVIDIA Omniverse on Instagram, Mediation And Twitter. For more join Inclusive society And check out the Omniverse Forums, Discord server, Twitch And Youtube Channels.