Overview
========

This framework is made up of four layers and accessible as both a CLI and library. These layers are as follows:

- **World Simulation:** Using existing high-quality assets and rendering engines render ground truth RGB images, depth maps, segmentation maps, normal maps, etc. 

- **Interpolation:** Using simulated data or data from an existing dataset optionally interpolate it to yield higher framerate datasets. This step can greatly help reduce the computational cost of emulating high speed sensors.

- **Sensor Emulation:** Apply realistic sensor modeling to the ground truth data to emulate different sensor modalities such as single photon cameras (both passive and active), event cameras and IMUs. 

- **Data Format and Loading:** Finally, all this data must be stored alongside all applicable metadata in a way that enables easy iteration and random access which is crucial for any deep learning applications.

.. TODO: Add examples for these use-cases

These layers can be used independently and as needed, making them very flexible. For instance, a VFX artist can use the world simulation layer to render out animations faster than what Blender can do by itself, or a computer vision researcher can uplift existing datasets, adding new sensor modalities to them by simply skipping the first step.