https://github.com/Erkaman/regl-cnn

Digit recognition with Convolutional Neural Networks in WebGL
https://github.com/Erkaman/regl-cnn

cnn convolutional-neural-networks deep-learning demo digit-recognition glsl gpgpu gpu javascript regl webgl

Last synced: about 1 month ago
JSON representation

Digit recognition with Convolutional Neural Networks in WebGL

• Host: GitHub
• URL: https://github.com/Erkaman/regl-cnn
• Owner: Erkaman
• License: mit
• Created: 2016-08-17T05:54:22.000Z (about 8 years ago)
• Default Branch: gh-pages
• Last Pushed: 2016-08-17T07:08:16.000Z (about 8 years ago)
• Last Synced: 2024-02-15T00:36:16.171Z (7 months ago)
• Topics: cnn, convolutional-neural-networks, deep-learning, demo, digit-recognition, glsl, gpgpu, gpu, javascript, regl, webgl
• Language: JavaScript
• Homepage: https://erkaman.github.io/regl-cnn/src/demo.html
• Size: 533 KB
• Stars: 502
• Watchers: 22
• Forks: 67
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-regl - GPU accelerated handwritten digit recognition with regl using Convolutional Neural Networks
• starred-awesome - regl-cnn - Digit recognition with Convolutional Neural Networks in WebGL (JavaScript)

README

        
# regl-cnn

GPU accelerated handwritten digit recognition with [regl](https://github.com/mikolalysenko/regl).

[Demo here](https://erkaman.github.io/regl-cnn/src/demo.html)

![Animated](gifs/record_resize.gif)

# Implementation Details

This demo does handwritten digit recognition by evaluating a

Convolutional Neural Network on the GPU with WebGL. The network was

trained in TensorFlow [by this script](https://github.com/Erkaman/regl-cnn/blob/gh-pages/scripts/create_cnn.py), and the network was then

reimplemented on the GPU by hand with WebGL. The main purpose of the

demo was to demonstate how our WebGL framework

[regl](https://github.com/mikolalysenko/regl) can be used to greatly

simplify GPGPU programming in WebGL. The secondary purpose was to

test whether evaluating Deep Learning networks in WebGL is doable.  To

our knowledge, our implementation is the first implementation ever to

attempt GPU accelerating neural networks with WebGL And we hope that

this implementation will provide a foundation for people who, like us,

wish to experiment with Deep Learning and WebGL The GPU implementation

can be found [here](https://github.com/Erkaman/regl-cnn/blob/gh-pages/src/gpu.js)

Note that this network will probably be slower than the corresponding

network implemented on the CPU. This is because of the overhead

associated with transferring data to and from the GPU. But in the

future we will attempt implementing more complex networks in the browser,

such as [Neural Style](https://arxiv.org/pdf/1508.06576v2.pdf), and then we think that we will see a

significant speedup compared to the CPU.

# Build

```bash

npm install

```

To then run the demo, do

```bash

npm run start

```

To run the test cases, do

```bash

npm run test

```