convnetjs README

The project's README from github.com/karpathy/convnetjs. As one markdown file: README.md.

ConvNetJS

ConvNetJS is a Javascript implementation of Neural networks, together with nice browser-based demos. It currently supports:

For much more information, see the main page at convnetjs.com

Note: I am not actively maintaining ConvNetJS anymore because I simply don't have time. I think the npm repo might not work at this point.

Online Demos

Example Code

Here's a minimum example of defining a 2-layer neural network and training it on a single data point:

// species a 2-layer neural network with one hidden layer of 20 neurons
var layer_defs = [];
// input layer declares size of input. here: 2-D data
// ConvNetJS works on 3-Dimensional volumes (sx, sy, depth), but if you're not dealing with images
// then the first two dimensions (sx, sy) will always be kept at size 1
layer_defs.push({type:'input', out_sx:1, out_sy:1, out_depth:2});
// declare 20 neurons, followed by ReLU (rectified linear unit non-linearity)
layer_defs.push({type:'fc', num_neurons:20, activation:'relu'}); 
// declare the linear classifier on top of the previous hidden layer
layer_defs.push({type:'softmax', num_classes:10});

var net = new convnetjs.Net();
net.makeLayers(layer_defs);

// forward a random data point through the network
var x = new convnetjs.Vol([0.3, -0.5]);
var prob = net.forward(x); 

// prob is a Vol. Vols have a field .w that stores the raw data, and .dw that stores gradients
console.log('probability that x is class 0: ' + prob.w[0]); // prints 0.50101

var trainer = new convnetjs.SGDTrainer(net, {learning_rate:0.01, l2_decay:0.001});
trainer.train(x, 0); // train the network, specifying that x is class zero

var prob2 = net.forward(x);
console.log('probability that x is class 0: ' + prob2.w[0]);
// now prints 0.50374, slightly higher than previous 0.50101: the networks
// weights have been adjusted by the Trainer to give a higher probability to
// the class we trained the network with (zero)

and here is a small Convolutional Neural Network if you wish to predict on images:

var layer_defs = [];
layer_defs.push({type:'input', out_sx:32, out_sy:32, out_depth:3}); // declare size of input
// output Vol is of size 32x32x3 here
layer_defs.push({type:'conv', sx:5, filters:16, stride:1, pad:2, activation:'relu'});
// the layer will perform convolution with 16 kernels, each of size 5x5.
// the input will be padded with 2 pixels on all sides to make the output Vol of the same size
// output Vol will thus be 32x32x16 at this point
layer_defs.push({type:'pool', sx:2, stride:2});
// output Vol is of size 16x16x16 here
layer_defs.push({type:'conv', sx:5, filters:20, stride:1, pad:2, activation:'relu'});
// output Vol is of size 16x16x20 here
layer_defs.push({type:'pool', sx:2, stride:2});
// output Vol is of size 8x8x20 here
layer_defs.push({type:'conv', sx:5, filters:20, stride:1, pad:2, activation:'relu'});
// output Vol is of size 8x8x20 here
layer_defs.push({type:'pool', sx:2, stride:2});
// output Vol is of size 4x4x20 here
layer_defs.push({type:'softmax', num_classes:10});
// output Vol is of size 1x1x10 here

net = new convnetjs.Net();
net.makeLayers(layer_defs);

// helpful utility for converting images into Vols is included
var x = convnetjs.img_to_vol(document.getElementById('some_image'))
var output_probabilities_vol = net.forward(x)

Getting Started

A Getting Started tutorial is available on main page.

The full Documentation can also be found there.

See the releases page for this project to get the minified, compiled library, and a direct link to is also available below for convenience (but please host your own copy)

Compiling the library from src/ to build/

If you would like to add features to the library, you will have to change the code in src/ and then compile the library into the build/ directory. The compilation script simply concatenates files in src/ and then minifies the result.

The compilation is done using an ant task: it compiles build/convnet.js by concatenating the source files in src/ and then minifies the result into build/convnet-min.js. Make sure you have ant installed (on Ubuntu you can simply sudo apt-get install it), then cd into compile/ directory and run:

$ ant -lib yuicompressor-2.4.8.jar -f build.xml

The output files will be in build/

Use in Node

The library is also available on node.js:

  1. Install it: $ npm install convnetjs
  2. Use it: var convnetjs = require("convnetjs");

License

MIT

convnetjs (karpathy/convnetjs)

Searching Database…

11.2K ·GitHub
Convnetjs
What is this?ConvNetJS is a JavaScript implementation of neural networks that lets you train Convolutional Neural Networks, ordinary fully-connected nets, and even reinforcement learning agents directly in the browser.
Is this a browser library or a Node package?Both: it runs client-side in the browser via the compiled convnet.js/convnet-min.js files, and it's also published as a Node.js module you can require() in server-side code.
How do I add it to a project?In Node, run npm install convnetjs and require it; in the browser, grab the minified build from the releases page and include convnet.js or convnet-min.js, then follow the Getting Started tutorial on the main site for example code.
Is this project still maintained?No — the author explicitly says he no longer has time to maintain it, and even warns that the npm package listing may be broken.
Who is this useful for today?It's mainly of historical/educational interest now for people wanting to see how a from-scratch JS neural net library works, rather than for production deep learning — modern projects should look at maintained libraries instead.
Where does the training actually run?Everything runs client-side in JavaScript — in the browser or under Node.js — with no server or Python backend involved.
What kinds of networks can ConvNetJS actually build?Fully connected layers with non-linearities, SVM/Softmax classification, L2 regression, full Convolutional Networks for image data, and an experimental Deep Q-Learning reinforcement learning module.
Are there live demos to try before installing anything?Yes — the README links browser demos including CNN training on MNIST digits and CIFAR-10, toy 2D/1D regression, an MNIST autoencoder, a Deep Q-Learning RL demo, and an image regression 'Painting' demo.
Where is the full documentation?Both the Getting Started tutorial and the full documentation live on the project's main site, convnetjs.com, rather than in the repo itself.
What are similar projects?brain.js, deep-learning-for-image-processing, face-api.js, Deep-Learning-Papers-Reading-Roadmap, conv_arithmetic, cs231n.github.io, fashion-mnist, deeplearning-models
Which topics is it in?convolutional-neural-networks, deep-reinforcement-learning, classification, reinforcement-learning, neural-network, deprecated, computer-vision, browser, deep-learning, machine-learning, education, npm
Creator
karpathy
Size
27.2 MB
Stars
11.2K
Updated
CommitNov 25Release2014.08.31Sep 01
Age
12y
Commits · 90 Days