Vector Drawing

Vector Data

Last week, we worked directly with pixels, the fundamental unit of raster graphics. Today we will work with vector graphics, which represent images as collections of shapes. Scalable Vector Graphics, or svg, is a common XML-based vector graphics format. SVG files look like this:

<?xml version="1.0" encoding="UTF-8" ?>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 360 144">
    <rect fill="#092" y="89.43" width="360" height="54.57"/>
    <rect fill="#0af" width="360" height="90"/>
    <rect fill="#fff" x="252" y="54" width="36" height="36"/>
    <polygon fill="#f00" points="288 54 270 36 252 54 288 54"/>
</svg>

Vector graphics must be converted to raster graphics—a process called rasterization—in order to be viewed on a pixel-based display. Fortunately, it is easy to rasterize vector graphics: vector graphics contain all the information necessary to do so. It is not easy to convert from raster images to vector images: In order to do so we would have to make inferences about the meaning implied by the raster image. When the form is ambiguous our inferences are just guesses, and they may be wrong.

Meaning vs. Form

Consider the image below. What would the image look like if the black circle were removed?

When we look at the image, we quickly infer that it represents two overlapping circles on a blue rectangle, but the raster image doesn’t contain information about shapes at all. It contains only blue, red, and black pixels. Humans are very, very good at inferring meaning from visual forms. We fill in meaning without conscious thought. When we look at the image above we see the circles automatically.

Because we are so good at inferring meaning, meaning is always there. We don’t appreciate how limited we would be without it. Computers are not good at inferring meaning from form. A human can guess that a red circle lies behind the black one, but a computer can’t—so the human can imagine the drawing without the circle, and the computer can’t.

This is the key advantage of vector graphics over raster graphics: vector graphics contain high-level meaning about the image that they represent. They are built from semantic units like rectangles and ellipses instead of nonsemantic pixels. This meaning allows us to make high-level changes: we can scale the vector image up and perfectly fill in the needed additional detail; we can change the fonts used to render text; we can remove a shape and reveal what is behind it.

::: .callout

Actually, computers can infer meaning from images, they just need to be programmed to do so. This is an extremely active area of research, and the rate of progress in recent years is staggering. This research is already making its way to consumer tools—Photoshop introduced Content-Aware Fill in 2010—and current approaches are actually better than humans at specific tasks.

The Two Minute Papers channel on YouTube summarizes graphics and AI research papers. Here are some great videos:

• Two Minute Papers: AI Learns Semantic Image Manipulation
• Two Minute Papers: Physics-based Image and Video Editing
• Two Minute Papers: Deep Image Prior
• Two Minute Papers: Image Colorization with Deep Learning and Classification
• Two Minute Papers: Learning to Fill Holes in Images /::

A Hybrid Workflow

Using a vector-based graphics library allows us to computationally generate images, export them in a way that preserves their shape information, and further manipulate them in vector editing software.

Paper.js

This chapter introduces Paper.js, developed by Jürg Lehni and Jonathan Puckey.

Paper.js is an open source vector graphics scripting framework that runs on top of the HTML5 Canvas. It offers a clean Scene Graph / Document Object Model and a lot of powerful functionality to create and work with vector graphics and bezier curves, all neatly wrapped up in a well designed, consistent and clean programming interface.

About Paper.js

You can find an overview of Paper.js features, a set of tutorials, and a complete API Reference on the Paper.js site.

PaperScript vs. JavaScript

Paper.js is a JavaScript library, and can be used with JavaScript alone, but can also be used with PaperScript. Paper.js is easier to use with PaperScript, at least for small projects. You can find info on setting up your workspace here: Working with Paper.js

You can use PaperScript with this site’s code example editor. You can tell the editor you are using PaperScript with a special comment: // language paperscript.

Here is an example of a Paper.js program written in PaperScript:

Library Exports

If you use Paper.js as a Javascript library, you will access the library through the paper object. In PaperScript, the library contents are exported to the scope of the program and can be accessed directly.

Javascript:

var path = new paper.Path();

PaperScript:

var path = new Path();

Operator Overloads

PaperScript also lets you use mathematical operators to work with Point and Size objects. You can see this on line 12. circle1Location + new Point(100, 0) adds two points together. JavaScript is not able to overload operators like this: this is a feature of PaperScript.

Let Does Not Work in PaperScript

Javascript now allows you to declar variables with let and const, but PaperScript uses only var

Retained Mode vs. Immediate Mode

The p5.js library is an immediate mode graphics library. When you call rect() or ellipse(), p5.js immediately draws the shape to the canvas.

Paper.js is a retained mode graphics library. When you call Path.Rectangles() or Path.Ellipses(), Paper.js creates object instances to represent the shapes and adds them to a scene graph. The scene graph can then be changed. You can add and remove shapes and change their properties. The shapes in the scene graph are all drawn to the canvas at once, when you are done making changes to it.

Creating an animation in p5.js involves constructing and drawing the entire canvas every frame. Paper.js redraws the shapes in your scene graph every frame, so all you have to do is change the properties of your animated shapes.

Notice that we don’t need to create the rectangle or the ellipse in every onFrame() call. We don’t need to mention the background rectangle at all, once it is in the scene graph. Paper.js will draw it every frame until we remove it. We also don’t need to set the color of the circle every frame. We can set it just when it changes, and that information is also stored in the scene graph.

SVG Output

Since Paper.js works in retained mode and stores your drawing as a vector scene graph, you can export your scene as an svg drawing that you can open in a vector editor like Adobe Illustrator.

The Project.exportSVG() function will create svg data of your Paper.js project. You need to do a little more work to download this data as a file. You can use the downloadAsSVG() in the example below to export your sketches. To export the example click it to focus it, and then press the s key.

Basic Example

Boolean Example

A super-useful feature of Paper.js is the ability to combine shapes with boolean operations. This allows you to describe complex shapes by building them from primitives.

Project Case Study

Paper Garden is a program written in about 300 lines of PaperScript using Paper.js. It generates stylized drawings of plants arranged in a wall garden. Read about how this program works and explore the code.

Paper Garden Case Study

Reference Links

YesYesNo: Más Que la Cara

An interactive installation in downtown Houston using paper.js and facial detection.

Autodraw

A Google AI experiment that interprets sketches, using Paper.js for the draw area.

Plumin.js

Javascript typeface editing using Paper.js.

Variable Fonts

Experiment with OpenType customizable typefaces.