Our tools will let you go beyond simulating how a 2D image will look when captured by a camera or other optical system. We've extended our very popular ISETcam to provide a physically-accurate simulation of how a 3D scene will look when renderd.
The tools in this repository calculate the spectral irradiance of realistic three-dimensional scenes. The tools work with a special version of PBRT (pbrt-v3-spectral) that we implemented. The implementation is available in a Docker container.
To use these tools you must have Docker and Matlab installed. This repository also depends on ISETCam (or ISETBio), and some people in our lab also require isetcloud. More on that in the wiki.
The general approach is the following
- Begin with a set of PBRT files that define the scene. We typically create scenes via the PBRT export tool from Cinema4D or more recently Blender.
(We have also built an extensive set of special tools to create complex automotive scenes. They require Flywheel and scitran)
Then a typical workflow might be:
- The user builds a recipe for rendering the PBRT files into a spectral irradiance. The recipe is a Matlab class that specifies spatial resolution, number of rays, viewing distance, type of optics (pinhole, lens or light field microlens array)
- The recipe also specifies information about the lens (which can contain multiple elements, spherical and certain aspherical shapes) and microlens array on the film surface
To see some examples, have a look at the tutorial directory. If you want to read more, please look through the wiki pages
Note: This repository was formerly pbrt2iset, and before that we relied on RenderToolbox4.
ISET3d was originally developed by ImageEval and development is now led Brian Wandell's Vistalab group at Stanford University, along with co-contributors from other research institutions and industry.