A poetic machine-learning dataset made of 5,000 images of found pebbles. Mimicking the datasets used to train computer vision systems, this project is an intervention that highlights the things we ignore when recording and classifying the world through data. Primarily sited as a repository on the code-sharing site Github, this project is also shown as a print of all the pebbles in the dataset sorted by visual similarity and a video.

Created while a Visiting Fellow (King's College) and artist-in-residence (Computer Laboratory) at University of Cambridge.

A detail of the pebbles shown above.
A screenshot of the project repository.
The dataset in another form: a video showing one pebble morphing into its most-similar neighbor, a kind of hallucination through the space of the dataset.

All work on this site is licensed under a Creative Commons BY-NC-SA License. Feel free to use, but please let me know.