Skip to content

beginaid/VAE

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

25 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

VAE

This repository is for sharing the scripts of VAE.

Latent space of the test-set data.

Description

You can train MNIST with any number of latent variable dimensions, and visualize the latent space from multiple perspectives.

Requirement

  • Python: 3.7
  • fire: 0.4

Usage

You can use the following command to train and visualize VAE.

python main.py main --z_dim [The dimensions of latent variables]

You can start tensorboard with the following command. Note that The default URL is http://localhost:6006/.

tensorboard --logdir ./logs

Network

The configuration of VAE is shown as below.

The configuration of VAE.

Output

You can get the following images.

Reconstruction

Reconstructions of the test-set data.

Mappings of the latent space

Mappings of the latent space in the test-set data.

Generation from lattice points

Artificially generated lattice points.


Reconstructions generated from lattice points.

Walkthrough

Four-direction-walkings at a constant speed in potential space.


Reconstructions obtained from a walkthrough of the latent space.