Proudly rendered with pytracer, head model courtesy of C348b.

PyTracer is a photo-realistic rendering interface backended by ray tracing. It may sound counter-intuitive that one may ever wish to implement a ray tracer with Python, nevertheles, PyTracer makes it feasible to do fast prototyping using current state-of-the-art learning libraries which are mostly implemented in Python.

Indeed, albeit to being a little bit slow, PyTracer can do it pretty well. Its object-oriented nature and modular design enables easy experiementing of new algorithms. The development and implementation of PyTracer largely take references from Physically Based Rendering, both second and third edition.

Currently, the integration with C++ has not completed, hence no compile is needed. PyTracer has the following dependencies:

• scipy: for KDTree structure for fast measured texture retreival
• numpy, PILLOW: for image I/O
• quaternion: for interpolation of motion transformations

To install the dependencies using pip, one may issue the following command:

pip install -r requirements.txt
pip install numpy-quaternion

After preparation, clone PyTracer via:

git clone https://github.com/zjiayao/pyTracer
cd pyTracer

One may also wish to unit test several low-level modules before use (assume pytest is installed, otherwise, one may do a pip install pytest):

PYTHONPATH=$PWD:$PYTHONPATH py.test

It might be convenient to move the pyTracer directory to your favourite location and add it to your PYTHONPATH (for example, under a virtualenv), assuming current directory is the root of pyTracer, one may add it via:

export PYTHONPATH=$PWD:$PYTHONPATH

The main work flow of PyTracer is analogous to pbrt. To start, one may render the sample head model image via:

python -c "import examples/head_mesh_render.py"

This gives a quick rendering of 128 * 128 pixels, with straitified sampling of 1 * 1 sampling rate. By default the rendered image is written to tmp.png.

To play around with other componenets, please visit the quick tutorial with examples at Quick Start Guide.

Currently, PyTracer supports the following features:

• Triangle Mesh with Loop Subdividing Surface Modeling;
• Full Spectrum Rendering (Sampling Spectrums Supported);
• Bounding Volume Hierarchy Accelerator;
• Projective, Perspective and Orthographic Cameras;
• Multiple Types of Textures;
• Multiple Types of Materials including Irregular Sampled Isotropic BRDF;
• Spot and Area Diffuse Lights;
• Monte Carlo Integration with MCMC Samplers;
• (Single) Path Integrator and Direct Lighting Integrator.

Users familiar with pbrt may find it intuitive to work with other components.

PyTracer is still under development for supporting more features, which are, tentatively:

• MERL BRDF Support;
• Volume Scattering Modeling;
• Bidirectional Path Tracing and More Light Transport Algorithm;
• Direct support to pbrt Flavour Input Files and Other UI/UX Improvements;
• Optimization with C++17;

The general goals for this stage are:

• Speed. PyTracer is currently amazingly slow.
• Robustness. Some components are implemented but have not been thoroughly tested yet.

PyTracer is maintained by Jiayao Zhang advising by Li-Yi Wei. The bib entry for this repo may be as follows:

@misc{pytracer:2017,
	title = {pyTracer},
	year = {2017},
	author = {Jiayao, Zhang and Li-Yi, Wei},
	publisher = {GitHub},
	journal = {GitHub Repository},
	howpublished= {\url{https://github.com/zjiayao/pyTracer}
}