I built a distributed ray tracer using C++ and created a 10-second animation.

The animation includes diffuse and Phong shading, soft shadows, and texture mapping. I implemented a simple BVH (bounding value hierarchy) structure to accelerate ray intersection, supersampling for anti-aliasing, and an object oriented approach to organize my code. 

Final project for CPSC 478 at Yale, Introduction to Computer Graphics

I began this project by pulling a figure from the CMU Motion Capture Database and a provided stick figure class. I chose a figure that was pretending to be a dinosaur and reoriented my camera to face the stick figure.

Motion capture stick figure, rendered in multiple spheres

I then placed the stick figure on a ground and added surrounding objects. The stick figure’s body was then changed from a series of spheres to consisting of cylinders with rounded ends. It was at this stage that I reorganized my code to be object oriented, which then allowed me to build out a more complex scene and tinker with ray-cylinder intersection.

Stick figure rendered in cylinders, standing on a ground and surrounded by castle-like scene

While implementing more complex and time intensive effects, I had to strike a balance between the scene I wanted to depict and the cost of rendering time. While I implemented Perlin noise, Oren-Nayer reflectance, and Cook-Torrance reflectance on parts of my scene, I realized that my scene would work best with texture mapping and soft shadows. I included Phong shading and an area light to mimic a sunset. While I considered keeping Cook-Torrance reflectance on the castle gate bars to add a more metallic effect, I decided to leave it out to reduce my render time.

The most challenging part of this project was implementing BVH. It required me to refactor a lot of code (which also solved a few bugs) and think critically about the structure of a BVH.

Screen capture from one of 300 final frames of the animation. 

I would do this project all over again in a heartbeat! But I would do a few things differently. I would first plan out my scene a little better from a creative standpoint to avoid having to implement many effects that I ended up not using. I would also test my BVH structure earlier and further develop into a subdivision that utilizes surface area heuristic. Lastly, I would render my video in a higher quality.

This was one of my favorite projects and I look forward to making more interesting animations with my newfound tools and knowledge.


GIF of the 10-second animation