Network View Crosshatcher
Houdini/Python
A side project I worked on while studying at Lost Boys. I used Houdini to process an image, and then used Houdini's Python API to re-create the cross-hatched image in the network view. The test subject for this example happens to be Harrison Molling, who was one of the TAs at Lost Boys during my time there.
Bumpy Dim
Python, Tensorflow, SMPL/STAR
A final project for a Deep Learning course at Brown. I worked with Anna Swanson and Thomas Jenkins on implementing a pose estimation model that would generate a 3D mesh based on a single-image input. We based our model off of TexturePose: Supervising Human Mesh Estimation with Texture Consistency written by Pavlakos et al. and End-to-end Recovery of Human Shape and Pose by Kanazawa et al. Our results were not quite on par with the papers we were implementing due to us training for a significantly shorter amount of time. Nonetheless, we saw some mesh/image imitation. Full writeup available here: https://devpost.com/software/bumpy-dim
GLSL/C++/Shadertoy
GLSL City
A procedurally generated city, my final in a graphics course. Geometry is generated through a ray-marcher written for an earlier assignment. All textures are generated procedurally. As this was part of a larger project, only the parts that I worked on are included in the reel. Some fine tuning of building heights and street widths were carried out by group member Dylan Tian. The shots in the video were made using a large window on a slow machine with playback speed adjusted. Frame rate on a medium-sized window is 15-25 FPS depending on machine horsepower. This component of the project is available on Shadertoy for public viewing: shadertoy.com/view/ltyfDd (Uses Web GL)