1. Getting Started With Replicator¶
Replicator (omni.replicator) is an Omniverse extension that provides various means of generating physically accurate synthetic data. To get a good understanding of its capabilities it is recommended to go over its documentation and tutorials.
1.1. The Semantics Schema Editor¶
To fully use the annotation capabilities of Replicator (e.g., semantic segmentation or 3d bounding boxes) the entities in the scene need to be semantically annotated. The Semantics Schema Editor extension is a GUI-based tool that can view, add, edit, or remove, class types or other key-value semantic information to prims in a stage.
To access the editor, navigate to Replicator > Semantics Schema Editor at the top of Kit. The editor will appear as a tab in the Properties pane by default.
For more information about the Semantics Schema Editor, please further refer to its documentation. For programmatically iterating and labeling a whole stage using various rules see the following example snippet as a starting point.
1.2. The Synthetic Data Visualizer¶
For a more thorough introduction to the visualizer tool, please further read the Visualizing Synthetic Data tutorial in the omni.replicator documentation.
1.2.1. Synthetic Data Visualizer Usage Example¶
Load the following warehouse stage that comes pre-labeled with semantic information:
In the Synthetic Data Visualizer, tick the sensors that you want to visualize
Press Visualize button to visualize output corresponding to the enabled sensors
1.3. The Synthetic Data Recorder¶
The Synthetic Data Recorder extension can be used as a GUI-based synthetic data acquisition tool built on top of Replicator’s API.
1.4. Additional Information¶
The session on Synthetic Data Generation using the Omniverse Replicator SDK from GTC 2022 is a good additional resource for a deeper understanding of Replicator.
Access to some GTC 2022 content is available only after completing the free registration to the NVIDIA Developer Network.