Things I've Made:

Chowder is a pair of glitch effects I made using the segmentation model FastSAM. Its name comes from the Cartoon Network show "Chowder," whose art style loosely inspired these effects.

The first effect, "reveal," simply segments the image and reveals a few of the segments each frame. The second effect, "scramble," replaces the texture of a few of the segments with that of a different segment. This was more difficult than I had initially imagined. Many of the segments are very irregular shapes, which meant I needed a way to stretch one irregular shape to another. To do this without too much math, I made two simplifying assumptions. First, I assumed that all of the segments were topologically identical. In other words, any holes in a segment would be ignored, and the shape would only be stretched based on its outer contour. Second, instead of stretching one shape to another, I would stretch each shape to a square texture image and subsequently sample from the second shape from that texture using a mask.

To do this, I found the 4 points which were closest to each of the bounding box's corners. The shape will be stretched such that these points are the 4 corners of the texture. Points on the contour between two corner points will be stretched to points evenly spaced along the corresponding edge of the texture. After creating the textures, each frame simply selects some number of segments to be modified and randomly selects a new texture for each one.

This project won't be on the "technology" side of this website, but you can find the code here: Github Repo.

In the process of building my Monte Carlo path tracer, I was testing different (incorrect) return values and took screenshots. These were my favorites.

This is a low resolution segmentation map from GraPL, the segmentation method I've been working on. The original image is a actually a man next to his drying fish, but the segmentation looks more pirate to me. BSDS is an incredibly interesting dataset to look through. They're all these old and colorful photos from around the world.

This image came about while I was debugging the face triangulation for UnShade, my facial shadow removal tool. I was trying to identify the indices of the triangles that comprise the eyebrows, eyes, nostrils, and mouth, since these pixels act as outliers in the inpainting calculations.