This project showcases a little CPU-based raytracer that can be used to render a single image, using a custom rendering library we entirely made from scratch, GLFW for presentation, ImGUI for UI and json lib for scene files.
The self-made custom library highlights our intermediate C++ knowledge by including many components implemented during the development of our project :
- Vector and 4x4 matrix operations, transform matrix management obviously as well as raycasting/ray intersection
- 7 basic 3-dimensional primitives have been implemented, from a basic triangle to a cone by way of an infinte cylinder
- Every object can have a unique ID'd material associated with a texture or a color
- Different illumination models (Phong and Lambert) and shadow casting in general are included
- The renderer supports skybox mapping and Super-Sampling Anti-Antialiasing x1-x16 (SSAA)
We also included a basic UI on top of the rendered scene to play with rendering parameters.
Since this project is currently mono-threaded, the drawing time approximately scales this way :
objects to render x lights x pixels to draw x subdivisions per pixel
As right now, there are no huge optimisation processes implemented.
As we did not really focus on the "beauty" aspect, please be lenient about the visual aspect of the rendered images.
Of course, there is a lot of room for improvement on this project, performance-wise and feature-wise.
Obviously, a major addition would be CPU multi-thread as it would speed up the drawing process.
A nice feature to add would be lens' aperture simulation / depth of field as it brings a nice aesthetic to rendered images.