topographytypography
Topography Typography is a project involving glyph detection and typeface creation from arbitrary satellite images. This project uses MNIST and Landsat data to train a neural network to detect physical glyphs and help curate a set of alphanumeric and punctuation forms for a given landscape.
This work has sparked several distinct field work projects including work in Northeast Peru, Bears Ears, Princeton, and Los Angeles. Download and read a collection of stories and technical explanations here.
This final paper served as my thesis project at Princeton University, and this work won Best Poster for Spring 2018 Princeton Computer Science independent research. I was graciously advised by several folks including David Reinfurt, David Dobkin, and Szymon Rusinkiewicz. Topography Typography is currently implemented using TensorFlow, Python, Illustrator, and Glyphs, and I plan to port this to ml5.js to more easily share this code.
2018—
