DCGAN in Tensorflow

Basic Implementation (Study friendly) of DCGAN in Tensorflow

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

• GAN: [Pytorch][Tensorflow]
• DCGAN: [Pytorch][Tensorflow]
• InfoGAN: [Pytorch][Tensorflow]
• Pix2Pix: [Pytorch][Tensorflow]
• DiscoGAN: [Pytorch][Tensorflow]

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)