Low-Code Workflows for Synthetic Data Generation Pipelines

With Omniverse Replicator, you can build custom synthetic data generation pipelines to create diverse training datasets with annotations such as depth, semantic and instance segmentation, and bounding boxes. In this demo, we use a simple YAML-based workflow to demonstrate how you generate training data by randomizing the location of objects such as a forklift, cones and wet floor signs, commonly found in a warehouse. Learn more about how you can bootstrap the training of your computer visions models with Omnivese Replicator.

-------------------------------------------------------

NVIDIA Omniverse is a scalable development platform for building and operating custom 3D pipelines and industrial metaverse applications. Based on Universal Scene Description (OpenUSD), Omniverse fundamentally transforms complex 3D workflows, allowing individuals and teams to connect disparate 3D tools and datasets, and simulate large-scale, physically accurate virtual worlds for industrial and scientific use cases.

Resources:

Connect with us on social:

#OpenUSD
#3D
#NVIDIAOmniverse

[sozno]

Nice video. 👍 The yaml addition is interesting. Would love a follow up video to go through the yaml file shown.