Performance-comparison-of-GAN-on-cifar-10

KO | EN

gitlite — search

#php #python #javascript #react #docker #redux #bot #css #vim #bootstrap #cpp #java

Performance-comparison-of-GAN-on-cifar-10

★ 37 Open GitHub ↗

Performance comparison of ACGAN, BEGAN, CGAN, DRAGAN, EBGAN, GAN, infoGAN, LSGAN, VAE, WGAN, WGAN_GP on cifar-10

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Performance-comparison-of-GAN-on-cifar-10
Performance comparison of ACGAN, BEGAN, CGAN, DRAGAN, EBGAN, GAN, infoGAN, LSGAN, VAE, WGAN, WGAN_GP on cifar-10

Reference：https://github.com/hwalsuklee/tensorflow-generative-model-collections 
The original code is for data MNIST, we changed the network structure to apply to cifar-10 and test Inception Score. 
The net structures are almost same. 
The following results can be reproduced with command:
>python main.py --dataset cifar-10 --gan_type <TYPE> --epoch 60 --batch_size 64

#ACGAN 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/ACGAN_IS.png" width="400">

#BEGAN 
The result is not well,we don't pay much time to to adjust the super-parameters. 
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/BEGAN_IS.png) 
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/BEGAN_train_59_0715.png)

#CGAN 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/CGAN_IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/CGAN_epoch059_test_all_classes.png)

#DRAGAN 
Stable, robust, fast convergent. 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/DRAGAN_epoch099IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/DRAGAN_epoch059_test_all_classes.png)

#EBGAN 
The net structure is the same as BEGAN, but collapse.

#GAN 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/GAN_epoch059IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/GAN_epoch059_test_all_classes.png)

#infoGAN 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/infoGAN_epoch059IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/infoGAN_epoch059_test_all_classes_style_by_style.png)

#LSGAN (Least Squares GAN) 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/LSGAN_epoch059IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/LSGAN_epoch059_test_all_classes.png)

#WGAN 
Not as well as paper. The net structure is the same as GAN, but converges too slowly.
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/WGAN_epoch059IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/WGAN_epoch059_test_all_classes.png)

#WGAN_GP 
There are total 300 epochs for Discriminator, but only 60 epochs for generator (The same times as other models). Converges slowly. 
<img src="https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/WGAN_GP_epoch299IS.png" width="400">
![](https://github.com/AliceAria/Performance-comparison-of-GAN-on-cifar-10/raw/master/images/WGAN_GP_epoch299_test_all_classes.png)

#VAE 
Collapsed. We also try to add or subtract bn layers, but it doesn't work.

Performance-comparison-of-GAN-on-cifar-10

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README