https://github.com/soumik12345/animegan

A Deep Convolutional Generative Adversarial Network to generate Faces of Anime Characters
https://github.com/soumik12345/animegan

anime convolutional-neural-networks dcgan dcgan-tensorflow face gan generative-adversarial-network keras python tensorflow

Last synced: 4 days ago
JSON representation

A Deep Convolutional Generative Adversarial Network to generate Faces of Anime Characters

• Host: GitHub
• URL: https://github.com/soumik12345/animegan
• Owner: soumik12345
• Created: 2019-05-16T14:19:24.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-05-16T15:03:37.000Z (over 5 years ago)
• Last Synced: 2024-12-27T14:27:09.138Z (14 days ago)
• Topics: anime, convolutional-neural-networks, dcgan, dcgan-tensorflow, face, gan, generative-adversarial-network, keras, python, tensorflow
• Language: Jupyter Notebook
• Size: 6.77 MB
• Stars: 11
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# AnimeGAN

A Deep Convolutional Generative Adversarial Network to generate Faces of Anime Characters

**Kaggle Notebook Link:** [https://www.kaggle.com/soumikrakshit/animegan](https://www.kaggle.com/soumikrakshit/animegan)

**Dataset Link:** [https://www.kaggle.com/soumikrakshit/anime-faces](https://www.kaggle.com/soumikrakshit/anime-faces)

## Dataset Sample

![Sample From the Dataset](./images/dataset-sample.png)

## AnimeGAN Architecture

AnimeGAN is based on the standard DCGAN architecture by Alec Radford, Luke Metz, Soumith Chintala. It mainly comprises of convolution layers without max pooling and fully connected layers. It uses convolutional stride and transposed convolution for the downsampling and the upsampling.

### Generator Architecture

![Generator Architecture](./images/generator.svg)

### Discriminator Architecture

![Discriminator Architecture](./images/discriminator.svg)

### Final Model

![Final Model Architecture](./images/gan.svg)

## Training AnimeGAN

The AnimeGAN was trained on 21551 anime face images from size `(64, 64)` and was trained for 15000 epochs for a batch size of 32 images using Nvidia Tesla K80 GPU in the Kaggle Kernel environment.

![Learning Faces](./images/training_process.gif)

## Training Results

![Learning Faces](./images/generator_loss.png)

![Learning Faces](./images/discriminator_loss.png)

![Learning Faces](./images/smaple_output.png)