Knot Classification was both a conceptual role testing and implementation prototype that used image classification to identify two different types of knots. I trained an image classification machine learning model on nearly 400 images of the monkey fist knot and trilobite paracord knot. The model successfully distinguishes the two different knot tying techniques. The idea for the prototype stemmed from the need to recognize both specific knot tying gesture sequences and the resulting knot. As a feedback to the maker, this presents an interesting opportunity explore the concept of human-machine symbiosis. Moving onwards, I would like to incorporate image classification into Unity and couple it with object recognition to try to detect the shape of a rope. Currently the prototype is written in JavaScript and uses the ML5.js library.
