Generative-models

Generative models using Keras and Pytorch

Requirement

• python3
• keras or pytorch

Keras

VAE

• parser description
  • -d : dataset name(mnist or fashion_mnist)
  • --latentdim : latent dimention(default=2)
  • --epoch : the number of epochs(default=2)
  • -b : the number of batch size(default=16)

Quick start

python3 ./VAE.py -d [dataset_name]

GAN

• parser description
  • -d : dataset name(mnist, fashio_manist or cifar10)
  • -b : batch size(default=128)
  • --epoch : the number of epochs(default=20)
  • --latentdim : latent dimention(default=100)
  • --leakyrelu : use leeaky-ReLU or not(default=True)
  • --tilt : tilt of leaky-ReLU(default=0.2)
  • --droprate : dropout rate(default=0.3)
  • --genoptim : generator's optimizer(default=adam)
  • --discoptim : discriminator's optimizer(default=adam)
  • --softlabel : use softlabel or not(default=False)

Quick start

python3 ./train.py -d [dataset_name]