Tutorials
Nine projects, each a self-contained Jupyter notebook.
Music Spirograph
Parametric geometry driven by audio features
A music-reactive spirograph renderer driven by audio analysis. Four moving entities trace synchronized geometric patterns using invisible circular colliders, producing smooth motion rendered frame-by-frame into video.
Music Painting
Mapping audio features to color and brushstroke via random walk
Maps spectral features extracted from music into continuously varying RGB colors via a custom mixing function. A random walk algorithm generates brush strokes whose width and direction shift with the underlying audio, producing paintings unique to each piece of music.
Neural Music Visualizer
CPPN-generated animation conditioned on audio features
Uses Compositional Pattern-Producing Networks (CPPNs) to create music-conditioned animated artwork. The network learns the color structure of a reference image and generates video frames whose visual character evolves with the audio.
Random Walk + Neural Networks
Random walk drawing with a coordinate network and audio modulation
Combines random-walk drawing with a coordinate-based neural network that learns the color structure of a reference image. Audio features modulate the walk's dynamics, producing animated paintings that blend the reference image's palette with the music's rhythm.
Neural Style Transfer
Separating content and style using VGG19 feature maps
Applies Neural Style Transfer using a pre-trained VGG19 network. Unlike typical deep learning, NST optimizes pixel values rather than weights — minimizing a combined content and style loss to produce images that carry one image's structure and another's texture.
Voronoi Photo Mosaic
Voronoi tessellation over sampled image points
Samples a set of points across an image, computes a Voronoi tessellation over them, and fills each cell with the average color of its underlying image region. The result is a geometric mosaic that preserves large-scale structure while abstracting fine detail.
Style Transfer Voronoi
Voronoi mosaic with neural style transfer applied
Combines Voronoi tessellation with Neural Style Transfer to produce stylized geometric mosaics. A photo is first decomposed into a Voronoi diagram whose cells are filled with averaged color, then Neural Style Transfer is applied to blend artistic texture into the geometric structure.
Algorithmic Photo Collage
Assembling a target image from a library of tile images
Algorithmically assembles photo collages by decomposing a target image and filling regions with best-matching tiles from an image library. Mathematical optimization governs placement, scale, and blending for a coherent result.
TSP Line Drawing
Image-to-point sampling and TSP optimization for single-line drawing
Converts an image into thousands of 2D points via density-driven sampling and Floyd-Steinberg dithering, then solves a large-scale Travelling Salesman Problem with the LKH-3 heuristic to produce a single unbroken line that renders the full image. SAM-based segmentation enables color-consistent multi-region drawings.