Skip to content
/ PGA Public

A TensorFlow implementation of perceptual generative autoencoder (PGA).

License

Notifications You must be signed in to change notification settings

zj10/PGA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PGA

A TensorFlow implementation of perceptual generative autoencoder (PGA), based on this implementation of VAE.

Usage

LPGA

# MNIST
train_lpga.py
# CIFAR-10
train_lpga.py --dataset cifar10 --model conv_32 --epoch 400 --lr 0.2 --bn True --z_dim 128 --zn_rec 3e-2
# CelebA
train_lpga.py --dataset celeba --model conv_64 --epoch 100 --lr 0.4 --bn True --z_dim 128 --zn_rec 3e-2 --zh_rec 1e-2 --nll 1e-2

VPGA

# MNIST
train_vpga.py
# CIFAR-10
train_vpga.py --dataset cifar10 --model conv_32 --epoch 400 --lr 0.2 --bn True --z_dim 128 --zn_rec 3e-2 --vrec 5e-3 --vkld 3e-2
# CelebA
train_vpga.py --dataset celeba --model conv_64 --epoch 60 --lr 0.4 --bn True --z_dim 128 --zn_rec 3e-2 --zh_rec 1e-2 --vkld 2e-3

LVPGA

# MNIST
train_lvpga.py
# CIFAR-10
train_lvpga.py --dataset cifar10 --model conv_32 --epoch 400 --lr 0.2 --bn True --z_dim 128 --zn_rec 3e-2 --vrec 1.5e-2 --vkld 5e-2 --nll 2e-2
# CelebA
train_lvpga.py --dataset celeba --model conv_64 --epoch 100 --lr 0.4 --bn True --z_dim 128 --zn_rec 3e-2 --zh_rec 1e-2 --vrec 1e-2 --vkld 5e-3 --nll 1e-2

About

A TensorFlow implementation of perceptual generative autoencoder (PGA).

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages