本文整理汇总了Python中torchvision.utils方法的典型用法代码示例。如果您正苦于以下问题：Python torchvision.utils方法的具体用法？Python torchvision.utils怎么用？Python torchvision.utils使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类torchvision的用法示例。
在下文中一共展示了torchvision.utils方法的4个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: get_test_imgs
# 需要导入模块: import torchvision [as 别名]
# 或者: from torchvision import utils [as 别名]
def get_test_imgs(args):
transform = torchvision.transforms.Compose([
torchvision.transforms.Resize((args.resize, args.resize)),
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
domA_test = CustomDataset(os.path.join(args.root, 'testA.txt'), transform=transform)
domB_test = CustomDataset(os.path.join(args.root, 'testB.txt'), transform=transform)
domA_test_loader = torch.utils.data.DataLoader(domA_test, batch_size=64,
shuffle=False, num_workers=0)
domB_test_loader = torch.utils.data.DataLoader(domB_test, batch_size=64,
shuffle=False, num_workers=0)
for domA_img in domA_test_loader:
if torch.cuda.is_available():
domA_img = domA_img.cuda()
domA_img = domA_img.view((-1, 3, args.resize, args.resize))
domA_img = domA_img[:]
break
for domB_img in domB_test_loader:
if torch.cuda.is_available():
domB_img = domB_img.cuda()
domB_img = domB_img.view((-1, 3, args.resize, args.resize))
domB_img = domB_img[:]
break
return domA_img, domB_img 示例2: cluster_acc
# 需要导入模块: import torchvision [as 别名]
# 或者: from torchvision import utils [as 别名]
def cluster_acc(Y_pred, Y):
from sklearn.utils.linear_assignment_ import linear_assignment
assert Y_pred.size == Y.size
D = max(Y_pred.max(), Y.max())+1
w = np.zeros((D,D), dtype=np.int64)
for i in range(Y_pred.size):
w[Y_pred[i], Y[i]] += 1
ind = linear_assignment(w.max() - w)
return sum([w[i,j] for i,j in ind])*1.0/Y_pred.size, w 示例3: removal
# 需要导入模块: import torchvision [as 别名]
# 或者: from torchvision import utils [as 别名]
def removal(args, e1, e2, d_a, d_b):
transform = torchvision.transforms.Compose([
torchvision.transforms.Resize((args.resize, args.resize)),
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
if args.eval_folder != '':
class Faces(data.Dataset):
"""Faces."""
def __init__(self, root_dir, transform, size, ext):
self.root_dir = root_dir
self.transform = transform
self.size = size
self.ext = ext
self.files = [f for f in os.listdir(root_dir) if f.endswith(ext)]
def __len__(self):
return self.size # number of images
def __getitem__(self, idx):
img_name = os.path.join(self.root_dir, self.files[idx])
image = Image.open(img_name)
sample = self.transform(image)
return sample
test_data = Faces(args.eval_folder, transform, args.amount, args.ext)
domA_test_loader = torch.utils.data.DataLoader(dataset=test_data, batch_size=args.bs, shuffle=False)
else:
domA_test = CustomDataset(os.path.join(args.root, 'testA.txt'), transform=transform)
domA_test_loader = torch.utils.data.DataLoader(domA_test, batch_size=args.bs, shuffle=False)
cnt = 0
for test_domA in domA_test_loader:
if torch.cuda.is_available():
test_domA = test_domA.cuda()
else:
test_domA = test_domA
test_domA = test_domA.view((-1, 3, args.resize, args.resize))
for i in range(args.bs):
separate_A = e2(test_domA[i].unsqueeze(0))
common_A = e1(test_domA[i].unsqueeze(0))
A_encoding = torch.cat([common_A, separate_A], dim=1)
A_decoding = d_a(A_encoding)
BA_decoding, mask = d_b(A_encoding, test_domA[i], A_decoding, args.threshold)
exps = torch.cat([test_domA[i].unsqueeze(0), BA_decoding], 0)
vutils.save_image(exps, '%s/%0d.png' % (args.out, cnt), normalize=True)
print(cnt)
cnt += 1
if cnt == args.amount:
break 示例4: __init__
# 需要导入模块: import torchvision [as 别名]
# 或者: from torchvision import utils [as 别名]
def __init__(self):
super(DCGAN_D, self).__init__()
main = torch.nn.Sequential()
### Start block
# Size = n_colors x image_size x image_size
if param.spectral:
main.add_module('Start-SpectralConv2d', torch.nn.utils.spectral_norm(torch.nn.Conv2d(param.n_colors, param.D_h_size, kernel_size=4, stride=2, padding=1, bias=False)))
else:
main.add_module('Start-Conv2d', torch.nn.Conv2d(param.n_colors, param.D_h_size, kernel_size=4, stride=2, padding=1, bias=False))
if param.SELU:
main.add_module('Start-SELU', torch.nn.SELU(inplace=True))
else:
if param.Tanh_GD:
main.add_module('Start-Tanh', torch.nn.Tanh())
else:
main.add_module('Start-LeakyReLU', Activation())
image_size_new = param.image_size // 2
# Size = D_h_size x image_size/2 x image_size/2
### Middle block (Done until we reach ? x 4 x 4)
mult = 1
ii = 0
while image_size_new > 4:
if param.spectral:
main.add_module('Middle-SpectralConv2d [%d]' % ii, torch.nn.utils.spectral_norm(torch.nn.Conv2d(param.D_h_size * mult, param.D_h_size * (2*mult), kernel_size=4, stride=2, padding=1, bias=False)))
else:
main.add_module('Middle-Conv2d [%d]' % ii, torch.nn.Conv2d(param.D_h_size * mult, param.D_h_size * (2*mult), kernel_size=4, stride=2, padding=1, bias=False))
if param.SELU:
main.add_module('Middle-SELU [%d]' % ii, torch.nn.SELU(inplace=True))
else:
if not param.no_batch_norm_D and not param.spectral:
main.add_module('Middle-BatchNorm2d [%d]' % ii, torch.nn.BatchNorm2d(param.D_h_size * (2*mult)))
if param.Tanh_GD:
main.add_module('Start-Tanh [%d]' % ii, torch.nn.Tanh())
else:
main.add_module('Middle-LeakyReLU [%d]' % ii, Activation())
# Size = (D_h_size*(2*i)) x image_size/(2*i) x image_size/(2*i)
image_size_new = image_size_new // 2
mult *= 2
ii += 1
### End block
# Size = (D_h_size * mult) x 4 x 4
if param.spectral:
main.add_module('End-SpectralConv2d', torch.nn.utils.spectral_norm(torch.nn.Conv2d(param.D_h_size * mult, 1, kernel_size=4, stride=1, padding=0, bias=False)))
else:
main.add_module('End-Conv2d', torch.nn.Conv2d(param.D_h_size * mult, 1, kernel_size=4, stride=1, padding=0, bias=False))
# Size = 1 x 1 x 1 (Is a real cat or not?)
self.main = main