Welcome to Limbo!
The Limbo project provides three things:
Limbo Data is a growing collection (~1000000 images and counting) of synthetic computer vision training data created for our research. If you’re a computer vision researcher interested in training models on synthetic data, we encourage you to take a look!
Limbo Data Format is the specification for how that data is stored. If you’re creating synthetic image data for research, we encourage you to consider adopting the same specification, and welcome feedback on how to improve it.
Limbo Software is a collection of Python-based tools for managing data in Limbo Data Format. Whether you’re using our Limbo Data or creating your own, these tools will simplify your work.