BigGAN-PyTorch

Pytorch implementation of LARGE SCALE GAN TRAINING FOR HIGH FIDELITY NATURAL IMAGE SYNTHESIS (BigGAN)

train imagenet

for 128*128*3 resolution

python main.py --batch_size 64  --dataset imagenet --adv_loss hinge --version biggan_imagenet --image_path /data/datasets

python main.py --batch_size 64  --dataset lsun --adv_loss hinge --version biggan_lsun --image_path /data1/datasets/lsun/lsun

python main.py --batch_size 64  --dataset lsun --adv_loss hinge --version biggan_lsun --parallel True --gpus 0,1,2,3 --use_tensorboard True

Different

Compatability

Pretrained Models

LSUN Pretrained model Download

Some methods in the paper to avoid model collapse, please see the paper and retrain your model.

Performance

Results

LSUN DATASETS(two classes): classroom and church_outdoor