https://github.com/rusty1s/embedded_gcnn

Embedded Graph Convolutional Neural Networks (EGCNN) in TensorFlow
https://github.com/rusty1s/embedded_gcnn

cnn gcnn graph neural-network spatial spectral tensorflow

Last synced: 2 months ago
JSON representation

Embedded Graph Convolutional Neural Networks (EGCNN) in TensorFlow

• Host: GitHub
• URL: https://github.com/rusty1s/embedded_gcnn
• Owner: rusty1s
• License: mit
• Created: 2017-03-01T13:45:14.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2018-06-10T07:31:51.000Z (over 6 years ago)
• Last Synced: 2024-11-29T18:11:49.098Z (2 months ago)
• Topics: cnn, gcnn, graph, neural-network, spatial, spectral, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 23.8 MB
• Stars: 78
• Watchers: 5
• Forks: 13
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Embedded Graph Convolutional Neural Networks

[![Build Status][build-image]][build-url]

[![Code Coverage][coverage-image]][coverage-url]

[![Requirements Status][requirements-image]][requirements-url]

[![Code Climate][code-climate-image]][code-climate-url]

[![Code Climate Issues][code-climate-issues-image]][code-climate-issues-url]

[build-image]: https://travis-ci.org/rusty1s/embedded_gcnn.svg?branch=master

[build-url]: https://travis-ci.org/rusty1s/embedded_gcnn

[coverage-image]: https://codecov.io/gh/rusty1s/embedded_gcnn/branch/master/graph/badge.svg

[coverage-url]: https://codecov.io/github/rusty1s/embedded_gcnn?branch=master

[requirements-image]: https://requires.io/github/rusty1s/embedded_gcnn/requirements.svg?branch=master

[requirements-url]: https://requires.io/github/rusty1s/embedded_gcnn/requirements/?branch=master

[code-climate-image]: https://codeclimate.com/github/rusty1s/embedded_gcnn/badges/gpa.svg

[code-climate-url]: https://codeclimate.com/github/rusty1s/embedded_gcnn

[code-climate-issues-image]: https://codeclimate.com/github/rusty1s/embedded_gcnn/badges/issue_count.svg

[code-climate-issues-url]: https://codeclimate.com/github/rusty1s/embedded_gcnn/issues

![Neural Network Approach](https://user-images.githubusercontent.com/6945922/28239620-a72734d0-6970-11e7-9253-8a1e027efd78.png)

This is a TensorFlow implementation of my mastersthesis on [Graph-based Image

Classification](https://github.com/rusty1s/deep-learning-on-graphs/tree/master/masterthesis)

*(german)*.

**Embedded graph convolutional neural networks (EGCNN)** aim to make significant improvements to learning on graphs where nodes are positioned on a twodimensional euclidean plane and thus possess an orientation (like up, down, right and left).

As proof, we implemented an image classification on embedded graphs by first segmenting the image into superpixels with the use of [SLIC](https://infoscience.epfl.ch/record/177415/files/Superpixel_PAMI2011-2.pdf) or [Quickshift](http://vision.cs.ucla.edu/papers/vedaldiS08quick.pdf), converting this representation into a graph and inputting these to the neural network.

![SlIC and Quickshift Segmentation](https://user-images.githubusercontent.com/6945922/27761633-61569a56-5e60-11e7-96d6-5a0507d26cf8.jpg)

Graphs are trained on three different datasets and are automatically downloaded by running the corresponding train scripts:

* [MNIST](http://yann.lecun.com/exdb/mnist/) (run `python mnist_graph.py` and `python mnist_spatial.py`)

* [Cifar-10](https://www.cs.toronto.edu/~kriz/cifar.html) (run `python cifar_graph.py` and `python cifar_conv2d.py`)

* [PascalVOC](http://host.robots.ox.ac.uk/pascal/VOC/) (run `python pascal_graph.py` and `python pascal_conv2d.py`)

This repository also includes layer implementations of alternative approaches such as [SGCNN](https://arxiv.org/abs/1312.6203) and [GCN](https://arxiv.org/abs/1609.02907) for graphs and the Fire module of [SqueezeNet](https://arxiv.org/abs/1602.07360) for images to validate the results.

## Results

| Dataset    | SLIC   | Quickshift |

| ---------- | ------:| ----------:|

| MNIST      | 97.405 | 98.025     |

| Cifar-10   | 74.218 | 75.230     |

| Pascal VOC | 54.473 | 54.516     |

## Requirements

To install the required python packages, run:

```bash

pip install -r requirements.txt

```

## Running tests

Install the test requirements

```bash

pip install -r requirements_test.txt

```

and run the test suite:

```bash

nosetests --nologcapture

```

## Cite

Please cite my master thesis if you use this code in your own work:

```

@mastersthesis{Fey2017,

  title={{Convolutional Neural Networks auf Graphrepr{\"a}sentationen von Bildern}},

  author={Matthias Fey},

  school={Technische Universit{\"a}t Dortmund},

  year={2017},

}

```