GAN

Conditional Image Generation with SaBN

Environment

• Ubuntu 16.04
• NVIDIA GPU
• python >= 3.6

Quick start

Installation:

1. Install pytorch >= v1.1.0 following official instruction.
2. Clone this repo:

git clone https://github.com/VITA-Group/Sandwich-Batch-Normalization
cd GAN

3. Install dependencies:

pip install -r requirements.txt

4. Prepare dataset:

   1. Please download ILSVRC2012 dataset from here.
   2. Follow the guideline here to preprocess the ImageNet dataset.

5. Download pretrained model weights from our model zoo (GoogleDrive) and put it to zoo.

mkdir zoo

Usage

Training

Experiments on CIFAR-10:

bash scripts/train_autogan_ccbn_cifar10.sh
bash scripts/train_autogan_sabn_cifar10.sh

bash scripts/train_sngan_ccbn_cifar10.sh
bash scripts/train_sngan_sabn_cifar10.sh

Experiments on ImageNet (cats and dogs):

bash scripts/sngan_ccbn_imagenet.sh
bash scripts/sngan_sabn_imagenet.sh

Testing

Tensorboard

tensorboard --logdir output --port 6001

Testing using model zoo's model weights (GoogleDrive)

bash scripts/test_autogan_sabn_cifar10.sh
bash scripts/test_sngan_sabn_cifar10.sh
bash scripts/test_sngan_sabn_imagenet.sh

Results

Evaluation results:

Model	Inception Score ↑	FID ↓
AutoGAN	8.43	10.51
BigGAN	8.91	8.57
SNGAN	8.76	10.18
AutoGAN-SaBN (ours)	8.72 (+0.29)	9.11 (−1.40)
BigGAN-SaBN (ours)	9.01 (+0.10)	8.03 (−0.54)
SNGAN-SaBN (ours)	8.89 (+0.13)	8.97 (−1.21)

Visual results on ImageNet (128*128 resolution):

SNGAN	SNGAN-SaBN (ours)