This is a implement of the paper 'Globally and Locally Consistent Image Completion'[Project Page]
For ATR Dataset:
CUDA_VISIBLE_DEVICES=0 python train --config=config.yaml
For LIP Dataset:
CUDA_VISIBLE_DEVICES=1 python train --config=config_lip.yaml