DCGAN pytorch CIFAR10

Implement of DCGAN pytorch using CIFAR10
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks, ICLR 2016

Train

python dcgan.py --dataroot [DATAROOT] --dataset [CIFAR] --model [DCGAN]

Configures

  model: dcgan
  is_train: True
  dataroot: dataset/cifar
  dataset: cifar
  download: True
  epochs: 25
  batch_size: 128
  image_size: 64
  nc: 3
  nz: 100
  ngf: 64
  ndf: 64
  learning_rate: 0.0002
  beta1: 0.5
  ngpu: 1
  cuda: True
  load_D: False
  load_G: False
  workers: 2
  generator_iters: 10000
  gpuids: [0]

Results

• generated images

• logs

References

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• https://github.com/pytorch/examples/tree/master/dcgan
• https://github.com/Zeleni9/pytorch-wgan
• https://github.com/sbarratt/inception-score-pytorch
• https://github.com/mseitzer/pytorch-fid