-
TensorFLow implementation of Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks which is a stabilize GAN by certain architectural constraints
-
Architecture guidelines for stable Deep Convolutional GANs:
- Replace any pooling layers with strided convolutions(discriminator) and fractional-strided convolutions(generator)
- Use BatchNorm in both the generator and the discriminator
- Remove fully connected hidden layers for deeper architectures
- Use Relu activation in generator for all layers except for the output, which uses Tanh
- Use LeakyReLU activation in the discriminator for all layers
-
Generator architecture
-
Usage
- Download dataset CelebA from here
- Train the model:
python main.py
-
Training details
-
To ensure thatd_loss
do not go to zero during the training, I run generator 5 times every discriminator updating. Although thed_loss
will not be close to 0, the quality of the generated images is very bad. -
Because I do not have GPUs to train all the training set now, so I just test the code with several samples on my MacBook Air. More training results will be updated later. -
Result
-
The code borrowed from here
-
Notifications
You must be signed in to change notification settings - Fork 0
huyz1117/DCGAN-TensorFlow
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
TensorFlow implementation of DCGAN
Topics
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published