Therefore, to the greatest degree possible, we must accept input geometry from any CAD environment. This is perhaps the most difficult of the goals we have outlined, as the detail and quality of CAD models varies widely. Many CAD systems and users produce only 2D or wireframe models, which are next to useless for simulation. Other CAD systems, capable of producing true 3D geometric models, cannot label the component surfaces and associate the material information necessary for an accurate lighting simulation. These systems require a certain degree of user intervention and post-processing to complete the model. Even the most advanced CAD systems, which produce accurate 3D models with associated surface data, do not break surfaces into meshes suitable for a radiosity calculation. The missing information must either be added by the user, inferred from the model, or the need for it must be eliminated. In our case, we eliminate this need by using something other than a radiosity (i.e. finite element) algorithm.
CAD translators have been written for AutoCAD, GDS, ArchiCAD, DesignWorkshop, StrataStudio, Wavefront, and Architrion, among others. None of these translators requires special intervention by the user to reorient surface normals, eliminate T-vertices, or mesh surfaces. The only requirement is that surfaces must somehow be associated with a layer or identifier that indicates their material type.