cpcdoy / cuda-pathtracer

Pathtracer using CUDA

Pathtracer with CUDA!

Simple Pathtracer made for our CUDA course at EPITA, a French Computer Science school.

Features

Scenes

In order to load a scene (or several at a time), you have to provide them on the command line. A scene is a special ASCII file having the following syntax:

cubemap cubemap/garden.jpg

# camera x y z dir_x dir_y dir_z fov_x dof_focus dof_aperture
camera -2.7 2.06 2.52 0.62 -0.348 -0.7 90.0 3.555 0.01

# pos_x pos_y pos_z r g b emission radius
p_light 2.9 2.1 2.9 1.0 1.0 1.0 2.0 0.3

# pos_x pos_y pos_z r g b emission radius
p_light -2.7 2.1 -2.55 1.0 1.0 1.0 2.0 0.3

# pos_x pos_y pos_z r g b emission radius
p_light 2.9 2.1 -2.55 1.0 1.0 1.0 2.0 0.3

scene obj/indoor.obj

Inside it, you have to reference the obj file you want to use, as well as a cubemap, some lights and a camera with some initial values.

Textures

We support the following textures:

• Diffuse
• Normal
• Specular
• Cubemaps

Algoritm

Our algorithm works using few samples, by using temporal buffering.

Build

Dependencies

You will need the following dependencies (fetched if using our CMake):

• GLFW

Windows

You will have to load the .sln, and go to the project properties window, in order to change the link to your include folders as well as the dependencies.

Linux

You can build the solution using:

sh$ mkdir build && cd build && cmake .. && make

Tool to visualize the sampling of our BRDF

At the root of the project, you can find a python Jupyter Notebook that helps visualize our BRDF sampling depending on roughness and such.