The concepts first
If you want to understand what engine-derived ground truth is before following a procedure, start with synthetic data generation in Unreal Engine.
Guides
How to actually do it.
Every guide here is written against the same real 40-frame capture, so the numbers in them are numbers you can check, and each one ends with something you can download rather than a call to action.
- How to generate synthetic data in Unreal EngineSeven steps from an existing level to a validated, exported dataset: classes, zones, population rules, validation, dry run, smoke test and generation.the whole workflow
- How to create a YOLO dataset in Unreal EngineWhat YOLO actually needs, how the export produces it, a real label file line by line, and the class balance to check before you start training.one format, in depth
Start somewhere else
The output first
If you would rather look at what comes out before reading how it is made, download a sample dataset and open it.
The reference documentation
Every Studio button, CLI flag, lens profile and weather profile is documented separately. Open the documentation.
More guides are being written, in the order people ask for them. If there is one you need, say so.