inpainting with pre-trained GAN in a feedback manner checkpoint: ./checkpoint test_img_path: ./test test_result_path: ./completions (automatically created) size:64 this can load either .jpg or .png images without modification trained in a identic mask: center, half ... which can be changed in function:complete in model.py the batchsize of test can be any int number