liujianzhao6328057/SGEN-pytorch

PyTorch implementation of "Sequential Gating Ensemble Network for Noise Robust Multi-Scale Face Restoration"

SGEN-pytorch

PyTorch implementation of "Sequential Gating Ensemble Network for Noise Robust Multi-Scale Face Restoration"

Requirements

• Python 2.7
• PyTorch
• torch-vision

Dataset

You first need to download the CelebA dataset from website (you're looking for a file called img_align_celeba.zip). Then, you need to create a folder structure as data/dataset_name/1.jpg,...,2.jpg,...

Training without GAN:

$ python main.py --dataset=dataset_name 
For example:
$ python main.py --dataset=img_align_celeba

Training with GAN:

$ python main.py --dataset img_align_celeba --is_trainwithGAN True

Testing:

$ python main.py --is_train False --dataset img_align_celeba_test --load_path logs/img_align_celeba_2018-08-13_13-50-12