1.Overview

This is the Pytorch implementation of variational auto-encoder, applying on MNIST dataset.

Currently, the following models are supported:

• :heavy_check_mark: VAE
• :heavy_check_mark: Conv-VAE

2.Usage

python train.py

The code is self-explanatory, you can specify some customized options in train.py.

3.Result

Here are some visualization results：

3.1 Reconstruction results

Model	epoch 10	epoch 20	epoch 30	epoch 40	epoch 50
VAE				| |
Conv-VAE

3.2 Randomly generated results

Model	epoch 10	epoch 20	epoch 30	epoch 40	epoch 50
VAE				| |
Conv-VAE

4. Pre-trained model

Donwload link: