https://github.com/andymatuschak/scrying-pen

A strange pen that draws in both the past and the future // a realtime implementation of SketchRNN
https://github.com/andymatuschak/scrying-pen

Last synced: 3 months ago
JSON representation

A strange pen that draws in both the past and the future // a realtime implementation of SketchRNN

• Host: GitHub
• URL: https://github.com/andymatuschak/scrying-pen
• Owner: andymatuschak
• License: other
• Created: 2018-03-04T01:19:14.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-03-27T04:45:57.000Z (over 6 years ago)
• Last Synced: 2024-07-18T05:38:37.776Z (4 months ago)
• Language: JavaScript
• Size: 703 KB
• Stars: 127
• Watchers: 7
• Forks: 18
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
# Scrying Pen

This pen's ink stretches backwards into the past and forwards into possible futures. The two sides make a strange loop: the future ink influences how you draw, which in turn becomes the new "past" ink influencing further future ink.

Put another way: this is a realtime implementation of [SketchRNN](https://arxiv.org/abs/1704.03477) which predicts future strokes while you draw.

[Play with it here!](http://andymatuschak.org/scrying-pen) (You'll need to use Chrome for now.)

![video of toy in action](https://andymatuschak.org/scrying-pen/images/hand.gif)

Let it guide your hand, or not, while you draw. Enjoy the gentle interplay between your volition and the machine's.

## Background

I'm excited about applications of machine learning to human augmentation—as opposed to automating tedious work or solving number-crunching problems. I believe effective media for human augmentation require feedback loops tight enough to work at the speed of thought. All that is to say: I'm interested in what happens when we take complex operations and make them interactive in real-time.

## Thanks

I'm grateful to David Ha, Douglas Eck, and their team for their great work with SketchRNN. The paper's wonderful to read; the models from Quick, Draw! are a great resource to the community; the demo implementations were a very helpful reference.

I'm also grateful to Michael Nielsen, Robert Ochshorn, and M Eifler for useful conversations about this project.

Thanks also to the [OpenAI Hackathon](https://blog.openai.com/hackathon/) for providing a nice venue for polishing the last bits of this project up.