A PyTorch implementation of Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
├── agents
| └── dcgan.py # the main training agent for the dcgan
├── graphs
| └── models
| | └── discriminator.py # discriminator model definition
| | └── generator.py # generator model definition
| └── losses
| | └── loss.py # contains the binary cross entropy
├── datasets # contains all dataloaders for the project
| └── celebA.py # dataloader for celebA dataset
├── data
| └── celebA # contains all celebA images
├── utils # utilities folder containing metrics , config parsing, etc
| └── assets
├── main.py
├── run.sh
CelebA dataset has been used. All images are resized to 64x64. Data are placed into data/ folder.
This is the model we used in our implementation. However, the number of filters in our implementation goes from 512 -> 256 -> 128 -> 64. This can be tuned in the configurations file be editing the variables: num_filt_g and num_filt_d.
- Input size: 64x64x3
- Batch size: 64
- Learning rate: 0.0002
- Betas for Adam: 0.5 and 0.999
- Number of epochs: 30
- Noise vector size: 100
- Starting number of generator filters: 64
- Starting number of discriminator filters: 64
- To run the project, you need to add your configurations into the folder configs/. An example of the configurations that should be passed can be found here
sh run.sh- To run on a GPU, you need to enable cuda in the config file.
Loss:
Generated Images after training:
- Pytorch: 0.4.0
- torchvision: 0.2.1
- tensorboardX: 1.2
Check requirements.txt.
- Pytorch official example: https://github.com/pytorch/examples/tree/master/dcgan
- DCGAN in torch: https://github.com/soumith/dcgan.torch
This project is licensed under MIT License - see the LICENSE file for details.