Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/mafda/generative_adversarial_networks_101

Keras implementations of Generative Adversarial Networks. GANs, DCGAN, CGAN, CCGAN, WGAN and LSGAN models with MNIST and CIFAR-10 datasets.
https://github.com/mafda/generative_adversarial_networks_101

ccgan ccgans cgan cgans cifar-10 cifar10 conda-environment dcgan gan gans generative-adversarial-network generative-adversarial-networks jupyter-notebook keras lsgan lsgans mnist mnist-dataset tensorflow wgan

Last synced: 27 days ago
JSON representation

Keras implementations of Generative Adversarial Networks. GANs, DCGAN, CGAN, CCGAN, WGAN and LSGAN models with MNIST and CIFAR-10 datasets.

Host: GitHub
URL: https://github.com/mafda/generative_adversarial_networks_101
Owner: mafda
License: mit
Created: 2018-06-14T13:43:36.000Z (about 6 years ago)
Default Branch: master
Last Pushed: 2023-11-08T01:10:52.000Z (8 months ago)
Last Synced: 2024-04-22T00:05:43.864Z (2 months ago)
Topics: ccgan, ccgans, cgan, cgans, cifar-10, cifar10, conda-environment, dcgan, gan, gans, generative-adversarial-network, generative-adversarial-networks, jupyter-notebook, keras, lsgan, lsgans, mnist, mnist-dataset, tensorflow, wgan
Language: Jupyter Notebook
Homepage:
Size: 16.1 MB
Stars: 199
Watchers: 8
Forks: 79
Open Issues: 0
Metadata Files:
- Readme: README.md
- License: LICENSE

Lists

awesome-stars - mafda/generative_adversarial_networks_101 - Keras implementations of Generative Adversarial Networks. GANs, DCGAN, CGAN, CCGAN, WGAN and LSGAN models with MNIST and CIFAR-10 datasets. (Jupyter Notebook)

README

        # Generative Adversarial Networks - GANs

This repository presents the **basic notions** that involve the concept of Generative Adversarial Networks.

> *"...the most interesting idea in the last 10 years in ML". Yann LeCun*

## Definition

[Generative Adversarial Networks or GANs](https://arxiv.org/abs/1406.2661) is a framework proposed by [Ian Goodfellow](http://www.iangoodfellow.com/), Yoshua Bengio and others in 2014.

GANs are composed of two models, represented by artificial neural network:

* The first model is called a **Generator** and it aims to generate new data similar to the expected one.

* The second model is named the **Discriminator** and it aims to recognize if an input data is ‘real’ — belongs to the original dataset — or if it is ‘fake’ — generated by a forger.



    



Read more in this post [GANs — Generative Adversarial Networks 101](https://mafda.medium.com/gans-generative-adversarial-networks-101-8bf8e304585c).

## Configure environment

- Create the conda environment

```shell

(base)$: conda env create -f environment.yml

```

- Activate the environment

```shell

(base)$: conda activate gans_101

```

- Run!

```shell

(gans_101)$: python -m jupyter notebook

```

## Models

Definition and training some models with MNIST and CIFAR-10 datasets.

### MNIST dataset

* GAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/01_GAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-generative-adversarial-network-with-mnist-part-db8b9c061de0)

* DCGAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/02_DCGAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-deep-convolutional-gans-with-mnist-part-3-8bad9a96ff65)

* CGAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/03_CGAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-conditional-gans-with-mnist-part-4-7f816d274d8c)

* CCGAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/04_CCGAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-context-conditional-gans-with-mnist-part-5-a8d56a243377)

* WGAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/05_WGAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-wasserstein-gan-with-mnist-part-6-7f796a0cea47)

* LSGAN - MNIST [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/06_LSGAN_MNIST.ipynb) - [Post Medium](https://mafda.medium.com/gans-least-squares-gans-with-mnist-part-7-e6cf2fff503c)

### CIFAR-10 dataset

* DCGAN - CIFAR10 [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/cifar10/02_DCGAN_CIFAR10.ipynb) - [Post Medium](https://mafda.medium.com/gans-deep-convolutional-gans-with-cifar10-part-8-be881a77e55b)

* CGAN - CIFAR10 [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/cifar10/03_CGAN_CIFAR10.ipynb) - [Post Medium](https://mafda.medium.com/gans-conditional-gans-with-cifar10-part-9-8e47373e33b8)

## Results

Training models with Keras - TensorFlow.

### MNIST dataset

#### Generative Adversarial Networks - GANs

A GANs implementation using fully connected layers. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/01_GAN_MNIST.ipynb)

| Epoch 00                          | Epoch 100                          | Loss                                |

| --------------------------------- | ---------------------------------- | ----------------------------------- |

| ![GAN with MNIST](img/00_gan.png) | ![GAN with MNIST](img/100_gan.png) | ![GAN with MNIST](img/loss_gan.png) |

#### Deep Convolutional Generative Adversarial Networks - DCGANs

A DCGANs implementation using the transposed convolution technique. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/02_DCGAN_MNIST.ipynb)

| Epoch 00                            | Epoch 100                            | Loss                                  |

| ----------------------------------- | ------------------------------------ | ------------------------------------- |

| ![GAN with MNIST](img/00_dcgan.png) | ![GAN with MNIST](img/100_dcgan.png) | ![GAN with MNIST](img/loss_dcgan.png) |

#### Conditional Generative Adversarial Nets - CGANs

A CGANs implementation using fully connected layers and embedding layers. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/03_CGAN_MNIST.ipynb)

| Epoch 00                            | Epoch 100                            | Loss                                  |

| ----------------------------------- | ------------------------------------ | ------------------------------------- |

| ![CGAN with MNIST](img/00_cgan.png) | ![CGAN with MNIST](img/100_cgan.png) | ![CGAN with MNIST](img/loss_cgan.png) |

#### Context-Conditional Generative Adversarial Networks - CCGANs

A CCGANs implementation using U-Net and convolutional neural network. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/04_CCGAN_MNIST.ipynb)

| Epoch 00                             | Epoch 100                             | Loss                                   |

| ------------------------------------ | ------------------------------------- | -------------------------------------- |

| ![CGAN with MNIST](img/00_ccgan.png) | ![CGAN with MNIST](img/100_ccgan.png) | ![CGAN with MNIST](img/loss_ccgan.png) |

#### Wasserstein Generative Adversarial Networks - WGANs

A WGANs implementation using convolutional neural network. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/05_WGAN_MNIST.ipynb)

| Epoch 00                            | Epoch 100                            | Loss                                  |

| ----------------------------------- | ------------------------------------ | ------------------------------------- |

| ![WGAN with MNIST](img/00_wgan.png) | ![WGAN with MNIST](img/100_wgan.png) | ![WGAN with MNIST](img/loss_wgan.png) |

#### Least Squares General Adversarial Networks - LSGANs

A LSGANs implementation using using fully connected layers. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/mnist/06_LSGAN_MNIST.ipynb)

| Epoch 00                                    | Epoch 100                                    | Loss                                          |

| ------------------------------------------- | -------------------------------------------- | --------------------------------------------- |

| ![LSGAN with MNIST](img/00_lsgan_mnist.png) | ![LSGAN with MNIST](img/100_lsgan_mnist.png) | ![LSGAN with MNIST](img/loss_lsgan_mnist.png) |

### CIFAR-10 dataset

#### Deep Convolutional Generative Adversarial Networks - DCGANs

A DCGANs implementation using the transposed convolution technique. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/cifar10/02_DCGAN_CIFAR10.ipynb)

| Epoch 00                                       | Epoch 100                                       | Loss                                             |

| ---------------------------------------------- | ----------------------------------------------- | ------------------------------------------------ |

| ![DCGAN with CIFAR-10](img/00_dcgan_cifar.png) | ![DCGAN with CIFAR-10](img/100_dcgan_cifar.png) | ![DCGAN with CIFAR-10](img/loss_dcgan_cifar.png) |

#### Conditional Generative Adversarial Networks - CGANs

A CGANs implementation using the transposed convolution and convolution neural network, and concatenate layers. [Notebook](https://github.com/mafda/generative_adversarial_networks_101/blob/master/src/cifar10/03_CGAN_CIFAR10.ipynb)

| Epoch 00                                     | Epoch 100                                     | Loss                                           |

| -------------------------------------------- | --------------------------------------------- | ---------------------------------------------- |

| ![CGAN with CIFAR-10](img/00_cgan_cifar.png) | ![CGAN with CIFAR-10](img/100_cgan_cifar.png) | ![CGAN with CIFAR-10](img/loss_cgan_cifar.png) |

---

## References

* Complete Post Medium

  * [GANs — Generative Adversarial Networks 101](https://mafda.medium.com/gans-generative-adversarial-networks-101-8bf8e304585c)

* Related papers:

    * [Generative Adversarial Networks](https://arxiv.org/abs/1406.2661)

    * [Unsupervised Representation Learning With Deep Convolutional](https://arxiv.org/pdf/1511.06434.pdf)

    * [Conditional Generative Adversarial Nets](https://arxiv.org/pdf/1411.1784.pdf)

    * [Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks](https://arxiv.org/pdf/1611.06430.pdf)

    * [Wasserstein GAN](https://arxiv.org/pdf/1701.07875.pdf)

    * [Least Squares General Adversarial Networks](https://arxiv.org/pdf/1611.04076.pdf)

* Datasets:

    * [THE MNIST DATABASE of handwritten digits](http://yann.lecun.com/exdb/mnist/)

    * [The CIFAR-10 dataset](https://www.cs.toronto.edu/%7Ekriz/cifar.html)

* Other repositories:

    * [Keras-GAN](https://github.com/eriklindernoren/Keras-GAN)

    * [Adversarial Nets Papers](https://github.com/zhangqianhui/AdversarialNetsPapers)

    * [How to Train a GAN? Tips and tricks to make GANs work](https://github.com/soumith/ganhacks)

    * [The GAN Zoo](https://github.com/hindupuravinash/the-gan-zoo)

    * [GAN Lab: An Interactive, Visual Experimentation Tool for Generative Adversarial Networks](https://github.com/poloclub/ganlab)

    * [gans-awesome-applications](https://github.com/nashory/gans-awesome-applications)

    * [tensorflow-generative-model-collections](https://github.com/hwalsuklee/tensorflow-generative-model-collections)

---

made with 💙 by [mafda](https://mafda.github.io/)