TinyPT

A tiny path tracer accelerated by OpenMP & CUDA.

Features

• Global illumination rendering with path tracing.
• Parallel rendering with OpenMP (on CPU) and CUDA (on nvidia GPU).
• Complex triangle mesh model from Wavefront OBJ with MTL texture, accelerated by BVH.
• Lambertian, metal, dielectric, and glossy (Cook-Torrance GGX) BxDFs.
• Alpha texture mapping, and bump texture with height map.
• Area light, point light, distant light, and environmental lighting.

Getting Started

Basic Usage

Firstly, clone the repo recursively to your local environment.

git clone --recursive https://github.com/li-plus/tinypt.git && cd tinypt
# or
git clone https://github.com/li-plus/tinypt.git && cd tinypt && git submodule update --init

Install compilation dependencies and download necessary scene resources.

sudo apt install gcc g++ cmake make libeigen3-dev libopencv-dev libgflags-dev libgoogle-glog-dev python3-dev python3-pip
bash resource/get_resource.sh

Build the project.

cmake -B build .
cmake --build build -j

Render an example scene with two spheres in a box, which is taken from smallpt.

./bin/pt -device cuda -num_samples 1000 -save_path scene.png -scene cornell_sphere

Python API

Python binding is also available for TinyPT. To build and install the Python package for development, run:

cd python/
python3 setup.py develop

Run a Python example.

python3 examples/main.py --device cuda --num-samples 1000 --save-path scene.png --scene cornell_sphere

To build the wheel for distribution, run:

python3 setup.py bdist_wheel

Docker

For those who do not have root privilege, we offer an alternative to build & run TinyPT within a docker container. For GPU support, nvidia-docker should be further installed.

cd docker
docker build . -t tinypt:latest --build-arg CUDA=1
bash run.sh tinypt:latest

More Examples

Following scenes are credited to smallpt and McGuire Computer Graphics Archive.

References

• smallpt [code]
• scratchapixel [blog]
• Ray Tracing in One Weekend [book] [code]
• Physically Based Rendering:From Theory To Implementation [book] [code]
• minpt [code]