https://github.com/mishig25/vizconvnets

Visualizing 2D Convolutional Layers
https://github.com/mishig25/vizconvnets

deeplearning tensorflow tensorflow-experiments tensorflowjs

Last synced: about 1 month ago
JSON representation

Visualizing 2D Convolutional Layers

• Host: GitHub
• URL: https://github.com/mishig25/vizconvnets
• Owner: mishig25
• License: mit
• Created: 2018-03-28T05:18:26.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2018-12-05T00:24:57.000Z (about 6 years ago)
• Last Synced: 2024-11-22T04:41:53.396Z (about 1 month ago)
• Topics: deeplearning, tensorflow, tensorflow-experiments, tensorflowjs
• Language: Jupyter Notebook
• Homepage: https://mishig25.github.io/vizconvnets/
• Size: 53.6 MB
• Stars: 6
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# VizConvnets

### Visualizing Channels of 2D Convolutional Layers through [Tensorflow.js](https://js.tensorflow.org)

Personal Project

#### Check out the Live Demo: [https://mishig25.github.io/vizconvnets/](https://mishig25.github.io/vizconvnets/)



#### Description:

After [AlexNet](https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf) won ImageNet 2012, popularity and usage of convnets have increased exponentially.

Visualizing channels/filters of conv layers turned out to be a powerful tool for analyzing how Convolutional Neural Networks work. [Zeiler and Fergus](https://arxiv.org/pdf/1311.2901.pdf) were one of the first people to visualize convnets throughly and went on to win ImageNet 2013.

Afterwards, there was a plethora of papers and demos about visualizing convnets, including [the popular one by Yosinski](http://yosinski.com/deepvis).

This project is a continuation of the convnet visualizing trend. By using Tensorflow.js and [MobileNet](https://arxiv.org/abs/1704.04861), an efficient CNN architecture, the project visualizes sample channels/filters from MobileNet and does so through web browser only.

### Contents of the repository:

* [Model](https://github.com/mishig25/vizconvnets/tree/master/model)

Activation model is created through Keras Functionall API in Jupyter Notebooks.

* [Frontend](https://github.com/mishig25/vizconvnets/tree/master/frontend)

Using Tensorflowjs and HTML5 Canvas to create a convnet visualizations in web-browser environemnt.

### Development:

```bash

git clone https://github.com/mishig25/vizconvnets.git

cd ./vizconvnets

cd frontend

yarn

yarn watch

```

#### License

MIT