https://github.com/devkihyun/vae-tensorflow

TensorFlow implementation of 'Variational Autoencoder'
https://github.com/devkihyun/vae-tensorflow

Last synced: 6 days ago
JSON representation

TensorFlow implementation of 'Variational Autoencoder'

• Host: GitHub
• URL: https://github.com/devkihyun/vae-tensorflow
• Owner: DevKiHyun
• Created: 2018-11-11T09:35:59.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2018-11-13T12:23:10.000Z (about 6 years ago)
• Last Synced: 2024-07-25T11:10:38.588Z (5 months ago)
• Language: Python
• Size: 270 KB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# VAE-Tensorflow (2018/11/11)

## Introduction

I implement a tensorflow model of a Variational Autoencoder for this paper[[Auto-Encoding Variational Bayes - by kingma]](https://arxiv.org/abs/1312.6114)

- I use mnist dataset as training dataset.

## Environment

- Ubuntu 16.04

- Python 3.5

## Depenency

- Numpy

- matplotlib

## Files

- vae.py : Model definition.

- main.py : Execute training and pass the default value.

- train.py : Training code.

## How to use

### Training

```shell

python main.py

# Default args: training_epoch = 200, z_dim = 20, batch_size = 128, learning_rate = 0.0001

# You can change args: training_epoch = 300, z_dim = 40 batch_size = 64, learning_rate = 0.0005

python main.py --training_epoch 300 --z_dim 40 --batch_size 64 --learning_rate 0.0005

```

## Result

### Reconstruction

![Alt Text](https://github.com/DevKiHyun/VAE-Tensorflow/blob/master/result/reconstruction.png)

### Generation

![Alt Text](https://github.com/DevKiHyun/VAE-Tensorflow/blob/master/result/generation.png)

### 2D-manifold

![Alt Text](https://github.com/DevKiHyun/VAE-Tensorflow/blob/master/result/manifold.png)

### 2D-manifold walking

![Alt Text](https://github.com/DevKiHyun/VAE-Tensorflow/blob/master/result/walking.png)

## Reference

[오토인코더의 모든것('All of autoencoder')](https://www.youtube.com/watch?v=o_peo6U7IRM&feature=youtu.be)

[Auto-Encoding Variational Bayes](https://arxiv.org/abs/1312.6114)