def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
transform = transforms.Compose([
transforms.Scale(opt.loadSize),
transforms.CenterCrop(opt.fineSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5),
(0.5, 0.5, 0.5))])
# Dataset A
dataset_A = ImageFolder(root=opt.dataroot + '/' + opt.phase + 'A',
transform=transform, return_paths=True)
data_loader_A = torch.utils.data.DataLoader(
dataset_A,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
# Dataset B
dataset_B = ImageFolder(root=opt.dataroot + '/' + opt.phase + 'B',
transform=transform, return_paths=True)
data_loader_B = torch.utils.data.DataLoader(
dataset_B,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
self.dataset_A = dataset_A
self.dataset_B = dataset_B
self.paired_data = PairedData(data_loader_A, data_loader_B)
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
transformations = [
#transforms.Scale(opt.loadSize),
MyScale(size=(256, 256), pad=True),
transforms.RandomCrop(opt.fineSize),
transforms.ToTensor(),
# this is wrong! because the fake samples are not normalized like this,
# still they are inferred on the same network,
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
#lambda x: (x - x.min()) / x.max() * 2 - 1, # [-1., 1.]
] # 归一化，会产生负数。
#transformations = [transforms.Scale(opt.loadSize), transforms.RandomCrop(opt.fineSize),
# transforms.ToTensor()]
transform = transforms.Compose(transformations)
# Dataset A, eg.. trainA目录
dataset_A = ImageFolder(root=opt.dataroot + '/' + opt.phase + 'A',
transform=transform,
return_paths=True)
data_loader_A = torch.utils.data.DataLoader(
dataset_A,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
# Dataset B
dataset_B = ImageFolder(root=opt.dataroot + '/' + opt.phase + 'B',
transform=transform,
return_paths=True)
data_loader_B = torch.utils.data.DataLoader(
dataset_B,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
# Dataset C
dataset_C = ImageFolder(root=opt.dataroot + '/' + opt.phase + 'C',
transform=transform,
return_paths=True)
data_loader_C = torch.utils.data.DataLoader(
dataset_C,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
# 如何保证 A、B、C是一一一对应的呢，shuffle=not self.opt.serial_batches 这个参数 serial_batches 为True，代表有序，否则随机
# shuffle 是洗牌，搅乱的意思
self.dataset_A = dataset_A
self.dataset_B = dataset_B
self.dataset_C = dataset_C
flip = opt.isTrain and not opt.no_flip
self.three_paired_data = ThreePairedData(data_loader_A, data_loader_B,
data_loader_C,
self.opt.max_dataset_size,
flip)
def __init__(self, opt):
super(ImageEdgeDataset, self).__init__(opt)
self.image_paths = ImageFolder(os.path.join(
self.root, opt.path_image)) # image path list
self.edge_paths = ImageFolder(os.path.join(
self.root, opt.path_edge)) # edge path list
assert len(self.image_paths) == len(self.edge_paths)
self.transform = get_transform()
def build_model(self):
""" DataLoader """
train_transform = transforms.Compose([
transforms.Resize((self.img_size, self.img_size)),
transforms.ToTensor(),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
])
test_transform = transforms.Compose([
transforms.Resize((self.img_size, self.img_size)),
transforms.ToTensor(),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
])
label_transform = transforms.Compose([
transforms.Resize((self.img_size, self.img_size)),
transforms.ToTensor()])
self.trainA = ImageFolder(os.path.join(self.dataset, 'trainA'), train_transform,extend_paths=True,return_paths=False)
self.trainB = ImageFolder(os.path.join(self.dataset, 'trainB'), train_transform,extend_paths=True,return_paths=False)
self.label = ImageFolder(os.path.join(self.dataset, 'label'), label_transform, extend_paths=True,loader="gray")
self.testA = ImageFolder(os.path.join(self.dataset, 'testA'), test_transform,extend_paths=True,return_paths=True)
self.testB = ImageFolder(os.path.join(self.dataset, 'testB'), test_transform,extend_paths=True,return_paths=True)
self.trainA_loader = DataLoader(self.trainA, batch_size=self.batch_size, shuffle=True,pin_memory=True)
self.trainB_loader = DataLoader(self.trainB, batch_size=self.batch_size, shuffle=True,pin_memory=True)
self.label_loader = DataLoader(self.label, batch_size=self.batch_size, shuffle=True,pin_memory=True)
self.testA_loader = DataLoader(self.testA, batch_size=1, shuffle=False,pin_memory=True)
self.testB_loader = DataLoader(self.testB, batch_size=1, shuffle=False,pin_memory=True)
""" Define Generator, Discriminator """
self.gen2B = networks.NiceResnetGenerator(input_nc=self.img_ch, output_nc=self.img_ch, ngf=self.ch, n_blocks=self.n_res, img_size=self.img_size, light=self.light).to(self.device)
self.gen2A = networks.NiceResnetGenerator(input_nc=self.img_ch, output_nc=self.img_ch, ngf=self.ch, n_blocks=self.n_res, img_size=self.img_size, light=self.light).to(self.device)
self.disA = networks.NiceDiscriminator(input_nc=self.img_ch, ndf=self.ch, n_layers=self.n_dis).to(self.device)
self.disB = networks.NiceDiscriminator(input_nc=self.img_ch, ndf=self.ch, n_layers=self.n_dis).to(self.device)
print('-----------------------------------------------')
input = torch.randn([1, self.img_ch, self.img_size, self.img_size]).to(self.device)
macs, params = profile(self.disA, inputs=(input, ))
macs, params = clever_format([macs*2, params*2], "%.3f")
print('[Network %s] Total number of parameters: ' % 'disA', params)
print('[Network %s] Total number of FLOPs: ' % 'disA', macs)
print('-----------------------------------------------')
_,_, _, _, real_A_ae = self.disA(input)
macs, params = profile(self.gen2B, inputs=(real_A_ae, ))
macs, params = clever_format([macs*2, params*2], "%.3f")
print('[Network %s] Total number of parameters: ' % 'gen2B', params)
print('[Network %s] Total number of FLOPs: ' % 'gen2B', macs)
print('-----------------------------------------------')
""" Define Loss """
self.L1_loss = nn.L1Loss().to(self.device)
self.MSE_loss = nn.MSELoss().to(self.device)
""" Trainer """
self.G_optim = torch.optim.Adam(itertools.chain(self.gen2B.parameters(), self.gen2A.parameters()), lr=self.lr, betas=(0.5, 0.999), weight_decay=self.weight_decay)
self.D_optim = torch.optim.Adam(itertools.chain(self.disA.parameters(), self.disB.parameters()), lr=self.lr, betas=(0.5, 0.999), weight_decay=self.weight_decay)
def CreateDataset(opt, test=False):
if opt.phase == 'train':
from data.image_folder import ImageFolder
target_dataset = ImageFolder(opt, opt.dataroot)
print("dataset was created")
return target_dataset
else:
from data.image_folder import ImageFolder
source_dataset = ImageFolder(opt, opt.dataroot)
print("dataset was created")
return source_dataset
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
self.fineSize = opt.fineSize
transformations = [
# TODO: Scale
transforms.Scale(opt.loadSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
]
transform = transforms.Compose(transformations)
# Dataset A
dataset = ImageFolder(root=opt.dataroot + '/' + opt.phase,
transform=transform,
return_paths=True)
data_loader = torch.utils.data.DataLoader(
dataset,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
self.dataset = dataset
flip = opt.isTrain and not opt.no_flip
self.paired_data = PairedData(data_loader, opt.fineSize,
opt.max_dataset_size, flip)
def get_data_loader_folder(input_folder,
batch_size,
train,
new_size=None,
height=256,
width=256,
num_workers=4,
crop=True):
transform_list = [
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
]
transform_list = [transforms.RandomCrop(
(height, width))] + transform_list if crop else transform_list
transform_list = [transforms.Resize(
new_size)] + transform_list if new_size is not None else transform_list
transform_list = [transforms.RandomHorizontalFlip()
] + transform_list if train else transform_list
transform = transforms.Compose(transform_list)
dataset = ImageFolder(input_folder, transform=transform)
loader = DataLoader(dataset=dataset,
batch_size=batch_size,
shuffle=train,
drop_last=True,
num_workers=num_workers)
return loader
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
self.fineSize = opt.fineSize
# transform the image data to the tensor range in (0, 1)
transform = transforms.Compose([
# this is fake, we will delete it later
# transforms.Scale((opt.loadSize, opt.loadSize), interpolation=Image.BILINEAR), # BICUBIC or ANTIALIAS
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5),
(0.5, 0.5, 0.5))])
# Dataset A
dataset = ImageFolder(root=opt.dataroot + '/' + opt.phase, caption = opt.caption, caption_bucket = opt.bucket_description, vocab = opt.vocab, data_augment = opt.augment_data,
transform=transform, return_paths=True)
data_loader = torch.utils.data.DataLoader(
dataset,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads),
drop_last=False
)
self.dataset = dataset
self.paired_data = PairedData(data_loader, opt.fineSize)
def __init__(self, params):
transform = transforms.Compose([
transforms.Scale(size=(params.load_size, params.load_size)),
transforms.RandomCrop(size=(params.height, params.width)),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
dataset_A = torch.utils.data.DataLoader(ImageFolder(
root=params.data_root + '/' + params.phase + 'A',
transform=transform),
num_workers=params.num_workers,
shuffle=params.shuffle)
dataset_B = torch.utils.data.DataLoader(ImageFolder(
root=params.data_root + '/' + params.phase + 'B',
transform=transform),
num_workers=params.num_workers,
shuffle=params.shuffle)
self.dataset_A = dataset_A
self.dataset_B = dataset_B
self.paired_data = PairedData(self.dataset_A, self.dataset_B)
def __init__(self, _root, _list_dir, _input_height, _input_width, _is_flip,
_shuffle):
transform = None
dataset = ImageFolder(root=_root, \
list_dir =_list_dir, input_height = _input_height, input_width = _input_width, transform=transform, is_flip = _is_flip)
data_loader = torch.utils.data.DataLoader(dataset,
batch_size=32,
shuffle=_shuffle,
num_workers=int(3))
self.dataset = dataset
flip = False
self.paired_data = PairedData(data_loader, flip)
def test(self):
args = self.args
input_dim = args.input_nc
if args.direction == 'AtoB':
a2b = 1
test_type = 'testA'
output_type = 'fakeB'
else:
a2b = 0
test_type = 'testB'
output_type = 'fakeA'
# Setup model and data loader
image_path = os.path.join(self.args.dataroot, test_type)
image_names = ImageFolder(image_path,
transform=None,
return_paths=True)
data_loader = get_data_loader_folder(image_path,
1,
False,
new_size=args.load_size,
crop=False)
checkpoint = os.path.join(args.result_dir, args.name, 'model',
'gen_%08d.pt' % (args.iteration))
try:
state_dict = torch.load(checkpoint)
self.gen_a.load_state_dict(state_dict['a'])
self.gen_b.load_state_dict(state_dict['b'])
except:
state_dict = pytorch03_to_pytorch04(torch.load(checkpoint))
self.gen_a.load_state_dict(state_dict['a'])
self.gen_b.load_state_dict(state_dict['b'])
self.to(self.device)
#self.eval()
encode = self.gen_a.encode if a2b else self.gen_b.encode # encode function
decode = self.gen_b.decode if a2b else self.gen_a.decode # decode function
for i, (images, names) in enumerate(zip(data_loader, image_names)):
print(names[1])
images = Variable(images.to(self.device), volatile=True)
content, _ = encode(images)
outputs = decode(content)
outputs = (outputs + 1) / 2.
# path = os.path.join(opts.output_folder, 'input{:03d}_output{:03d}.jpg'.format(i, j))
basename = os.path.basename(names[1])
path = os.path.join(args.result_dir, args.name, output_type,
basename)
if not os.path.exists(os.path.dirname(path)):
os.makedirs(os.path.dirname(path))
vutils.save_image(outputs.data, path, padding=0, normalize=True)
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
self.fineSize = opt.fineSize
transform = transforms.Compose([
# TODO: Scale
transforms.Resize(opt.loadSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
# Dataset A
dataset = ImageFolder(root=opt.dataroot + '/' + opt.phase,
transform=transform,
return_paths=True,
font_trans=(not opt.flat),
rgb=opt.rgb,
fineSize=opt.fineSize,
loadSize=opt.loadSize)
data_loader = torch.utils.data.DataLoader(
dataset,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
self.dataset = dataset
dict_inds = {}
test_dict = opt.dataroot + '/test_dict/dict.pkl'
print("test_dict", test_dict)
if opt.phase == 'test':
if os.path.isfile(test_dict):
with open(test_dict, 'rb') as f:
dict_inds = pickle.load(f, encoding='bytes')
else:
warnings.warn(
'Blanks in test data are random. create a pkl file in ~/data_path/test_dict/dict.pkl including predifined random indices'
)
if opt.flat:
self._data = FlatData(data_loader, data_loader_base, opt.fineSize,
opt.max_dataset_size, opt.rgb, dict_inds,
opt.base_font, opt.blanks)
else:
self._data = Data(data_loader, opt.fineSize, opt.max_dataset_size,
opt.rgb, dict_inds, opt.blanks)
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
self.fineSize = opt.fineSize
transform = transforms.Compose([
# TODO: Scale
#transforms.Scale((opt.loadSize * 2, opt.loadSize)),
#transforms.CenterCrop(opt.fineSize),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
# Dataset A
dataset = ImageFolder(root=opt.dataroot + '/' + opt.phase,
transform=transform,
return_paths=True)
data_loader = torch.utils.data.DataLoader(
dataset,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
self.dataset = dataset
self.paired_data = PairedData(data_loader, opt.fineSize)
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
transform = transforms.Compose([
# TODO: Scale
transforms.Resize(opt.loadSize),
transforms.ToTensor(),
# transforms.Normalize((0.5, 0.5, 0.5),
# (0.5, 0.5, 0.5))
])
dic_phase = {'train': 'Train', 'test': 'Test'}
# Dataset A
dataset_A = ImageFolder(root=opt.dataroot + '/A/' + opt.phase,
transform=transform,
return_paths=True,
rgb=opt.rgb,
fineSize=opt.fineSize,
loadSize=opt.loadSize,
font_trans=False,
no_permutation=opt.no_permutation)
len_A = len(dataset_A.imgs)
if not opt.no_permutation:
shuffle_inds = np.random.permutation(len_A)
else:
shuffle_inds = range(len_A)
dataset_B = ImageFolder(root=opt.dataroot + '/B/' + opt.phase,
transform=transform,
return_paths=True,
rgb=opt.rgb,
fineSize=opt.fineSize,
loadSize=opt.loadSize,
font_trans=False,
no_permutation=opt.no_permutation)
if len(dataset_A.imgs) != len(dataset_B.imgs):
raise Exception(
"number of images in source folder and target folder does not match"
)
if (opt.partial and (not self.opt.serial_batches)):
dataset_A.imgs = [dataset_A.imgs[i] for i in shuffle_inds]
dataset_B.imgs = [dataset_B.imgs[i] for i in shuffle_inds]
dataset_A.img_crop = [dataset_A.img_crop[i] for i in shuffle_inds]
dataset_B.img_crop = [dataset_B.img_crop[i] for i in shuffle_inds]
shuffle = False
else:
shuffle = not self.opt.serial_batches
data_loader_A = torch.utils.data.DataLoader(
dataset_A,
batch_size=self.opt.batchSize,
shuffle=shuffle,
num_workers=int(self.opt.nThreads))
data_loader_B = torch.utils.data.DataLoader(
dataset_B,
batch_size=self.opt.batchSize,
shuffle=shuffle,
num_workers=int(self.opt.nThreads))
if opt.base_font:
# Read and apply transformation on the BASE font
dataset_base = ImageFolder(root=opt.base_root,
transform=transform,
return_paths=True,
font_trans=True,
rgb=opt.rgb,
fineSize=opt.fineSize,
loadSize=opt.loadSize)
data_loader_base = torch.utils.data.DataLoader(
dataset_base,
batch_size=1,
shuffle=False,
num_workers=int(self.opt.nThreads))
else:
data_loader_base = None
self.dataset_A = dataset_A
self._data = PartialData(data_loader_A, data_loader_B,
data_loader_base, opt.fineSize, opt.loadSize,
opt.max_dataset_size, opt.phase,
opt.base_font)
# fineSize=176, gpu_ids=[0], half=False, how_many=200, init_type='normal',
# input_nc=1, isTrain=True, l_cent=50.0, l_norm=100.0, lambda_A=1.0, lambda_B=1.0,
# lambda_GAN=0.0, lambda_identity=0.5, loadSize=256, load_model=True, lr=0.0001,
# lr_decay_iters=50, lr_policy='lambda', mask_cent=0.5, max_dataset_size=inf,
# model='pix2pix', n_layers_D=3, name='siggraph_retrained',
# ndf=64, ngf=64, niter=100, niter_decay=100, no_dropout=False, no_flip=False,
# no_html=False, no_lsgan=False, norm='batch', num_threads=1, output_nc=2, phase='val',
# pool_size=50, print_freq=200, resize_or_crop='resize_and_crop',
# results_dir='./results/', sample_Ps=[1, 2, 3, 4, 5, 6, 7, 8, 9],
# sample_p=1.0, save_epoch_freq=1, save_latest_freq=5000, serial_batches=True,
# suffix='', update_html_freq=10000, verbose=False, which_direction='AtoB',
# which_epoch='latest', which_model_netD='basic', which_model_netG='siggraph')
dataset = ImageFolder(opt.dataroot,
transform=transforms.Compose([
transforms.Resize((opt.loadSize, opt.loadSize)),
transforms.ToTensor()
]))
dataset_loader = torch.utils.data.DataLoader(
dataset, batch_size=opt.batch_size, shuffle=not opt.serial_batches)
model = create_model(opt)
model.setup(opt)
model.eval()
# pdb.set_trace();
# create website
web_dir = os.path.join(opt.results_dir, opt.name,
'%s_%s' % (opt.phase, opt.which_epoch))
def CreateDataset():
dataset = None
from data.image_folder import ImageFolder
dataset = ImageFolder('../../dataset/sbu')
return dataset
def initialize(self, opt):
BaseDataLoader.initialize(self, opt)
transformations = [
#transforms.Scale(opt.loadSize),
MyScale(size=(256, 256), pad=True),
transforms.RandomCrop(opt.fineSize),
transforms.ToTensor(),
# this is wrong! because the fake samples are not normalized like this,
# still they are inferred on the same network,
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
#lambda x: (x - x.min()) / x.max() * 2 - 1, # [-1., 1.]
] # 归一化，会产生负数。
#transformations = [transforms.Scale(opt.loadSize), transforms.RandomCrop(opt.fineSize),
# transforms.ToTensor()]
transform = transforms.Compose(transformations)
# Dataset A1, eg.. train/A1目录
dataset_A1 = ImageFolder(root=opt.dataroot + '/' + opt.phase + '/A1',
transform=transform,
return_paths=True)
data_loader_A1 = torch.utils.data.DataLoader(
dataset_A1,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
dataset_A2 = ImageFolder(root=opt.dataroot + '/' + opt.phase + '/A2',
transform=transform,
return_paths=True)
data_loader_A2 = torch.utils.data.DataLoader(
dataset_A2,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
dataset_B1 = ImageFolder(root=opt.dataroot + '/' + opt.phase + '/B1',
transform=transform,
return_paths=True)
data_loader_B1 = torch.utils.data.DataLoader(
dataset_B1,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
dataset_B2 = ImageFolder(root=opt.dataroot + '/' + opt.phase + '/B2',
transform=transform,
return_paths=True)
data_loader_B2 = torch.utils.data.DataLoader(
dataset_B2,
batch_size=self.opt.batchSize,
shuffle=not self.opt.serial_batches,
num_workers=int(self.opt.nThreads))
# 如何保证 A1、A2、B1、B2是一一一对应的呢，shuffle=not self.opt.serial_batches 这个参数 serial_batches 为True，代表有序，否则随机
# shuffle 是洗牌，搅乱的意思
#
# 奇怪，如果A和B的数据数量不一致呢？也可以在opt里面修改load的数量大小的
self.dataset_A1 = dataset_A1
self.dataset_A2 = dataset_A2
self.dataset_B1 = dataset_B1
self.dataset_B2 = dataset_B2
flip = opt.isTrain and not opt.no_flip
self.four_paired_data = FourPairedData(data_loader_A1, data_loader_A2,
data_loader_B1, data_loader_B2,
self.opt.max_dataset_size, flip)