Skip to content

HyeongminLEE/Tensorflow_DCGAN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DCGAN in Tensorflow

Basic Implementation (Study friendly) of DCGAN in Tensorflow

[Paper | Post(in Korean) | Pytorch Version]

1. Environments

  • Windows 10
  • Python 3.5.3 (Anaconda)
  • Tensorflow 1.4.0
  • Numpy 1.13.1
  • lmdb (pip install lmdb): for LSUN Dataset
  • cv2 (conda install -c conda-forge opencv): for LSUN Dataset

2. Networks and Parameters

2.1 Hyper-Parameters

  • Image Size = 64x64 (Both in CelebA and LSUN-Bedroom)
  • Batch Size = 128 (~32 is OK)
  • Learning Rate = 0.0002
  • Adam_beta1 = 0.5
  • z_dim = 100
  • Epoch = 5 in CelebA is Enough, 1 in LSUN is Enough. Sometimes it can be diverge.

2.2 Generator Networks (network.py)

2.3 Discriminator Networks (network.py)

3. Run (Train)

You can modify hyper-parameter. Look at the parsing part of the code.

3. 1 CelebA DB (Cropped Face, 156253 Samples)

  • Database Setting: link

  • Train & Test

python train.py --filelist <filelist_name> --out_dir <output_directory>
  • Test results will be saved in 'output_directory'

3. 2 LSUN-Bedroom DB (3033042 Samples)

  • Database Setting: link

  • Train & Test

python train.py --filelist <filelist_name> --out_dir <output_directory>
  • Test results will be saved in 'output_directory'

4. Results

DCGAN with CelebA (6 Epochs)

DCGAN with LSUN (1 Epochs)

About

Study Friendly Implementation of DCGAN in Tensorflow

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages