if args.data == 'CelebA-HQ':
from data import CelebA_HQ
test_dataset = CelebA_HQ(args.data_path, args.attr_path, args.image_list_path, args.img_size, 'test', args.attrs)
os.makedirs(output_path, exist_ok=True)
test_dataloader = data.DataLoader(
test_dataset, batch_size=1, num_workers=args.num_workers,
shuffle=False, drop_last=False
)
if args.num_test is None:
print('Testing images:', len(test_dataset))
else:
print('Testing images:', min(len(test_dataset), args.num_test))
attgan = AttGAN(args)
attgan.load(find_model(join('output', args.experiment_name, 'checkpoint'), args.load_epoch))
progressbar = Progressbar()
attgan.eval()
for idx, (img_a, att_a) in enumerate(test_dataloader):
if args.num_test is not None and idx == args.num_test:
break
img_a = img_a.cuda() if args.gpu else img_a
att_a = att_a.cuda() if args.gpu else att_a
att_a = att_a.type(torch.float)
att_b_list = [att_a]
if args.by_levels:
test_dataset = CelebA(args.data_path, args.attr_path, args.img_size, 'mytest',
args.attrs)
test_dataloader = data.DataLoader(test_dataset,
batch_size=1,
num_workers=args.num_workers,
shuffle=False,
drop_last=False)
print('Testing images:', len(test_dataset))
output_path = join('output', args.experiment_name, 'attention_testing')
os.makedirs(output_path, exist_ok=True)
attgan = AttGAN(args)
attgan.load(r'weights_unzip.17.pth')
attgan.eval()
for idx, (img_real, att_org) in enumerate(test_dataloader):
img_real = img_real.cuda() if args.gpu else img_real
att_org = att_org.cuda() if args.gpu else att_org
att_org = att_org.type(torch.float)
_, mc, mw, mh = img_real.shape
att_list = [att_org]
img_unit = img_real.view(3, mw, mh)
img_unit = ((img_unit * 0.5) + 0.5) * 255
img_unit = np.uint8(img_unit)
img_unit = img_unit[::-1, :, :].transpose(1, 2, 0)
for i in range(args.n_attrs):
tmp = att_org.clone()
tmp[:, i] = 1 - tmp[:, i]
from os.path import join
from attgan import AttGAN
def parse(args=None):
parser = argparse.ArgumentParser()
parser.add_argument('--experiment_name',
dest='experiment_name',
type=str,
default='06-36AM on March 09, 2021')
parser.add_argument('--load_epoch', dest='load_epoch', type=int, default=0)
parser.add_argument('--gpu', dest='gpu', type=bool, default=False)
return parser.parse_args(args)
args_ = parse()
with open(join('output', args_.experiment_name, 'setting.txt'), 'r') as f:
args = json.load(f, object_hook=lambda d: argparse.Namespace(**d))
args.gpu = args_.gpu
args.experiment_name = args_.experiment_name
args.load_epoch = args_.load_epoch
args.betas = (args.beta1, args.beta2)
model = AttGAN(args)
model.load(
os.path.join('output', args.experiment_name, 'checkpoint',
'weights.' + str(args.load_epoch) + '.pth'))
model.saveG_D(os.path.join('output', args.experiment_name, 'checkpoint',
'weights_unzip.{:d}.pth'.format(args.load_epoch)),
flag='unzip')
batch_size=1,
num_workers=args.num_workers,
shuffle=False,
drop_last=False)
# 如果没有限定处理多少张，那就把整个数据集都做迁移
if args.num_test is None:
print('Testing images:', len(test_dataset))
else:
print('Testing images:', min(len(test_dataset), args.num_test))
# 载入AttGAN模型
attgan = AttGAN(args)
# 载入指定节点
attgan.load(
find_model(join('output', args.experiment_name, 'checkpoint'),
args.load_epoch)
) # 载入指定节点 output/128_shortcut1_inject1_none/checkpoint/
progressbar = Progressbar()
# 进行验证
attgan.eval()
# 对图片的大循环
for idx, (img_a, att_a) in enumerate(test_dataloader):
'''
idx: 图像的索引
img_a: 图像
att_a: 标签
原始标签 label_a
1.文件名 '{:06d}.jpg'.format(idx + 182638) name_array[i]
2.生成标签 att_c_list[i] att_b_list[i]
3.生成的图片 samples[i] samples[i]
'''