Diffy.js
A dependency-free motion detection library for the browser.
This library came out of my browser-based interactive experiments and from the need to extract motion data from the webcam through the getUserMedia
API. This is the cleaned up version of that code. Hat tip to Soundstep for the technique used in this library.
Overview
Basically, Diffy.js grabs two consecutive webcam snapshots in each tick of the loop (via requestAnimationFrame
) & combines them into a high contrast blended image to create a "diff image". This image can be adjusted from the API via sensitivity
and threshold
parameters. Based on a resolution: {x, y}
parameter from the API, Diffy.js will create a matrix containing average values from this image. This matrix is then passed as the only argument to a recursively-executed callback function provided by the user: onFrame: function(matrix) { /* draw something */ }
. A Web Worker is used to speed up image processing. A simple visual debugger can be optionally turned on as you experiment with values.
Screenshot from demo/
: Raw webcam input, Mirrored raw canvas, "diff" image canvas, a simple canvas experiment with Diffy.js
Installation
npm: npm install diffyjs --save
Bower: bower install diffyjs --save
Git: git clone https://github.com/maniart/diffyjs.git
Usage
With ES2015 via Babel:
; const diffy = ;
With ES5 via <script>
tag:
<!-- HTML: -->
// JS:var diffy = Diffy;
A few things to keep in mind
- Diffy.js is meant to be used in a Browser environment
- The library will request camera access upon loading of the web page. Choose
Allow
to proceed. - It also requires Web Workers.
- Any hosted project using the
getUserMedia
API (including Diffy.js), must be served overHTTPS
, but you can easily run it onlocalhost
. - Diffy.js is designed to allow 1 instance on each page. Further instantiation attempts will throw.
API Reference
#create(options)
Creates and returns a Diffy.js instance. It will request camera access as soon as the web page loads and will immediately begin executing the provided callback function.
Arguments
- options (object)
- resolution (object) [default:
{x: 10, y: 5}
] - defines the size of the output matrix- x (number) [default:
10
] - resolution along the X axis - y (number) [default:
5
] - resolution along the Y axis
- x (number) [default:
- sensitivity (number) [default:
0.2
] - a decimal value between 0 and 1. It impacts the contrast of the blended image. Somewhere around 0.2 is usually good. yay magic numbers! - threshold (number) [default:
21
] - any number between 0 and 255 can be used. But ahem magic numbers are around 20 and 25. Experiment with this. This parameter defines the minimum average value that registers as "movement" for Diffy.js) - debug (boolean) [default:
false
] - hides or shows the debug view. Please note that to work with video and pixel data, Diffy.js will add a few DOM nodes & a minimalstyle
tag to your page. However, they are hidden unlessdebug
flag is set totrue
. - sourceDimensions (object) [default:
{x: 130, h: 100}
] - defines the dimensions for the source frame. Keep in mind that the larger the dimensions, the more pixels Diffy.js needs to examine, hence the lower the performance.- x (number) [default:
130
] - width of the source frame - y (number) [default:
100
] - height of the source frame
- x (number) [default:
- containerClassName (string) [default:
diffy--debug-view
] - defines the class name for the container element that wraps around Diffy.js debug view. Debug view is hidden by default, unless thedebug
flag is set totrue
. - onFrame (function) [default:
() => {}
] - callback function executed recursively at each tick of the loop by Diffy.js viarequestAnimationFrame
API. Diffy.js provides this function with the motion matrix as the only argument.
- resolution (object) [default:
Development
This project uses Node.js v6.9.1
, Babel v6.18.0
and Webpack v1.13.2
.
After cloning the repo, install the dev dependencies from the project root via npm install
.
At this point, you should be able to run npm start
to start the development server at http://localhost:3000
.
npm start
will watch source files in src
for changes, recompiles the changes files and serves them from the memory. In order to produce build artifacts, run npm run build
. This will output the build files and source maps in dist/
and demo/dist
Demo
To run the demo included in the project:
- Run
npm install
to install the required npm packages. - Run
npm run demo
from the project root. This script will create build artifacts indist/
anddemo/dist
and start a demo server athttp://localhost:4000
. - Direct your browser to
http://localhost:4000
Tests
As of right now, test coverage is [___covers face___] embarrassingly low but you can run them via npm test
from the project root. New releases will include better test coverage.
Contribute
Contributions / comments much welcome! Open a Pull Request and lets work together?
Issues
Please report issues here.
License
MIT. See LICENSE