Introduction

It is a GAN embedded with CBAM and SENet. We trained it on mnist, fashion-mnist and a dataset of anime faces.

Performance

On mnist dataset:

On fashion-mnist dataset:

On anime faces dataset:

Unzip and move it into : ./data/dataset_anime/data/faces

Train the model:

nohup bash train.sh > ./logs/res.txt 2> ./logs/err.txt &

or:

python train.py --dataset mnist --gan_type gan --epoch 50 --batch_size 64 --use_cuda True

The result will be saved in ./results/