This repository has been archived by the owner. It is now read-only.
Run trained deep neural networks in the browser or node.js
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
build deps Feb 13, 2016
src remove JZS1/2/3 layers Feb 14, 2016
test remove JZS1/2/3 layers Feb 14, 2016
.babelrc defer implementation of useGPU flag for now Nov 17, 2015
.gitignore include browserify-built minified js file in repo Oct 14, 2015
.npmignore update npmignore Oct 14, 2015
LICENSE defer implementation of useGPU flag for now Nov 17, 2015


Run trained deep neural networks in the browser or node.js. Currently supports serialization from trained Keras models (note: not up-to-date with the current Keras API).

build status npm version


Training deep neural networks on any meaningful dataset requires massive computational resources and lots and lots of time. However, the forward-pass prediction phase is relatively cheap - typically there is no backpropagation, computational graphs, loss functions, or optimization algorithms to worry about.

What do you do when you have a trained deep neural network and now wish to use it to power a part of your client-facing web application? Traditionally, you would deploy your model on a server and call it from your web application through an API. But what if you can deploy it in the browser alongside the rest of your webapp? Computation would be offloaded entirely to your end-user!

Perhaps most users will not be able to run billion-parameter networks in their browsers quite yet, but smaller networks are certainly within the realm of possibility.

The goal of this project is to provide a lightweight javascript library that can take a serialized Keras, Caffe, Torch or [insert other deep learning framework here] model, together with pre-trained weights, pack it in your webapp, and be up and running. Currently supports serialization from trained Keras models.


  • MNIST multi-layer perceptron / src / demo

  • CIFAR-10 VGGNet-like convolutional neural network / src / demo

  • LSTM recurrent neural network for classifying astronomical object names / src / demo

You can also run the examples on your local machine at http://localhost:8000:

$ npm run examples-server


See the source code of the examples above. In particular, the CIFAR-10 example demonstrates a multi-threaded implementation using Web Workers.

In the browser:

<script src="neocortex.min.js"></script>
  // use neural network here

In node.js:

$ npm install neocortex-js
import NeuralNet from 'neocortex-js';

The core steps involve:

  1. Instantiate neural network class
let nn = new NeuralNet({
  // relative URL in browser/webworker, absolute path in node.js
  modelFilePath: 'model.json',
  arrayType: 'float64', // float64 or float32
  1. Load the model JSON file, then once loaded, feed input data into neural network
nn.init().then(() => {
  let predictions = nn.predict(input);
  // make use of predictions


To build the project yourself, for both the browser (outputs to build/neocortex.min.js) and node.js (outputs to dist/):

$ npm install
$ npm run build

To build just for the browser:

$ npm run build-browser


A script to serialize a trained Keras model together with its hdf5 formatted weights is located in the utils/ folder. It currently only supports sequential models with layers listed below. Implementation of graph models is planned.

Functions and layers currently implemented are listed below. More forthcoming.

  • Activation functions: linear, relu, sigmoid, hard_sigmoid, tanh, softmax

  • Advanced activation layers: leakyReLULayer, parametricReLULayer, parametricSoftplusLayer, thresholdedLinearLayer, thresholdedReLuLayer

  • Basic layers: denseLayer, dropoutLayer, flattenLayer, mergeLayer

  • Recurrent layers: rGRULayer (gated-recurrent unit or GRU), rLSTMLayer (long short-term memory or LSTM)

  • Convolutional layers: convolution2DLayer, maxPooling2DLayer, convolution1DLayer, maxPooling1DLayer

  • Embedding layers: embeddingLayer (maps indices to corresponding embedding vectors)

  • Normalization layers: batchNormalizationLayer


$ npm test

Browser testing is planned.


Thanks to @halmos for the logo.




Apache 2.0