Overview

Recognize drawn shapes based on their point clouds.

Demos:

• Point Cloud.
• Unistroke.

Installation

npm install outlines

Usage

var outlines = require('outlines'),
    recognizer = outlines.Recognizer(),
    Point = outlines.Point,
    gesture = [];

// Create an array of the points making up a gesture you want to find
// a match for.
// new Point(x, y, strokeId)
// strokeId should be incremented for each new stroke. Ie. If a user lifts and
// touches/mouses down again each subsequent stream of points is a separate stroke
gesture.push(new Point(0, 0, 0))
gesture.push(new Point(10, 10, 0))
gesture.push(new Point(20, 20, 0));

gesture.push(new Point(40, 40, 1));
gesture.push(new Point(50, 50, 1));
gesture.push(new Point(60, 60, 1));

// Get a list in ranked (closest to furthest match)
// Note: For now there is a built in set of symbols. You could replace a recognizers
// set of symbols though, see the implementation to figure out how:
// https://github.com/francoislaberge/outlines/blob/master/src/outlines.js#L107-L152
var matches = recognizer.Rank(gesture);
console.log(matches);

References

• Based on the $P Point-Cloud Recognizer research paper
  • $P project page
  • PDF of paper
  • Pseudo Code

Coming Soon

Support for more recognizers will be added soonish. Here's a sneak peak:

• Work in progress: Unistroke Recognizer.
• Want to add the GRAIL system to this as well
  • https://jackschaedler.github.io/handwriting-recognition/
  • See Video for motivation
• Talk to Ken Perlin and ask what he's using for Chalktalk
• Some notes related to the history of hand writing recognition
  • http://www.osnews.com/story/26838/Palm_I_m_ready_to_wallow_now/page2/

Experiments I'd like to make:

• Try reusing icon sets such as FontAwesome, as starting points.
  • Related: Allow SVG and fonts as potential sources?
    • For SVG: Try: http://helpfulsheep.com/2015-03-25-converting-svg-fonts-to-svg/