def transforms():
for i in tqdm(glob.glob(os.path.join(opt.train_data_root, '*.bmp'))):
uid = uuid.uuid1()
profix = "NEG"
img = Image.open(i)
if 'POS' in i.split('/')[-1]:
profix = "POS"
p = profix + "-" + str(uid) + "1.bmp"
data = T.RandomHorizontalFlip(p=1)(img)
data.save(os.path.join("./data/train", p))
p = profix + "-" + str(uid) + "2.bmp"
data = T.RandomVerticalFlip(p=1)(img)
data.save(os.path.join("./data/train", p))
p = profix + "-" + str(uid) + "3.bmp"
data = img.rotate(90)
data.save(os.path.join("./data/train", p))
p = profix + "-" + str(uid) + "4.bmp"
data = img.rotate(270)
data.save(os.path.join("./data/train", p))
p = profix + "-" + str(uid) + "5.bmp"
data = T.RandomRotation(90)(img)
data.save(os.path.join("./data/train", p))
p = profix + "-" + str(uid) + "6.bmp"
data = T.RandomRotation(270)(img)
data.save(os.path.join("./data/train", p))
def lfw(stage, configs=None, augment=False, tta=False, tta_size=48):
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
folder_path = os.path.join(configs["data_path"], f"{stage}")
dataset = datasets.ImageFolder(
folder_path,
transforms.Compose([
#transforms.RandomResizedCrop(224),
#transforms.RandomHorizontalFlip(),
transforms.Resize(224),
transforms.ToTensor(),
#normalize,
]),
)
if augment:
dataset2 = datasets.ImageFolder(
folder_path,
transforms.Compose([
transforms.Resize(224),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation((1, 20)),
transforms.ColorJitter(),
transforms.ToTensor(),
]),
)
dataset3 = datasets.ImageFolder(
folder_path,
transforms.Compose([
transforms.Resize(224),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation((-20, -1)),
transforms.ColorJitter(),
transforms.ToTensor(),
]),
)
dataset4 = datasets.ImageFolder(
folder_path,
transforms.Compose([
transforms.Resize(224),
transforms.RandomHorizontalFlip(),
transforms.RandomAffine(degrees=0, translate=(0.1, 0.1)),
transforms.ColorJitter(),
transforms.ToTensor(),
AddGaussianNoise(0., 0.05),
]),
)
return torch.utils.data.ConcatDataset(
[dataset4, dataset2, dataset3, dataset])
else:
return dataset
def get_mnist_loader(data_dir='dataset', batch_size=128, num_workers=4):
transform = transforms.Compose([
transforms.RandomRotation((-30, 30)),
transforms.ToTensor(),
])
return DataLoader(dataset=MNIST(root=data_dir, train=True, transform=transform, download=True), batch_size=batch_size,
shuffle=True, num_workers=num_workers)
def get_simclr_pipeline_transform(size, s=1, num_aug=5):
"""Return a set of data augmentation transformations as described in the SimCLR paper."""
color_jitter = transforms.ColorJitter(0.8 * s, 0.8 * s, 0.8 * s,
0.2 * s)
if num_aug == 5:
data_transforms = transforms.Compose([
transforms.RandomResizedCrop(size=size),
transforms.RandomHorizontalFlip(),
transforms.RandomApply([color_jitter], p=0.8),
transforms.RandomGrayscale(p=0.2),
GaussianBlur(kernel_size=int(0.1 * size)),
transforms.ToTensor()
])
elif num_aug == 7:
data_transforms = transforms.Compose([
transforms.RandomResizedCrop(size=size),
transforms.RandomHorizontalFlip(),
transforms.RandomApply([color_jitter], p=0.8),
transforms.RandomGrayscale(p=0.2),
GaussianBlur(kernel_size=int(0.1 * size)),
transforms.RandomRotation(degrees=45),
transforms.RandomAffine(degrees=45),
transforms.ToTensor()
])
return data_transforms
def get_transforms():
tsfm = []
if rcp.transforms.topilimage: tsfm += [transforms.ToPILImage()]
if rcp.transforms.randomrotation:
tsfm += [transforms.RandomRotation(rcp.transforms.randomrotation)]
if rcp.transforms.randomverticalflip:
tsfm += [
transforms.RandomVerticalFlip(rcp.transforms.randomverticalflip)
]
if rcp.transforms.randomhorizontalflip:
tsfm += [
transforms.RandomHorizontalFlip(
rcp.transforms.randomhorizontalflip)
]
if rcp.transforms.colorjitter:
tsfm += [transforms.ColorJitter(**rcp.transforms.colorjitter)]
# if rcp.transforms.randomcrop: tsfm += [transforms.RandomCrop(rcp.transforms.randomcrop)]
if rcp.transforms.resize:
tsfm += [
transforms.Resize((rcp.transforms.resize, rcp.transforms.resize))
]
if rcp.transforms.totensor: tsfm += [transforms.ToTensor()]
if rcp.transforms.normalize:
tsfm += [transforms.Normalize(**rcp.transforms.normalize)]
return transforms.Compose(tsfm)
def get_dataset_loaders(
dataset_directory,
minibatch_size) -> Tuple[DataLoader, DataLoader, DataLoader]:
data_transform = transforms.Compose([
transforms.RandomRotation(10),
transforms.Resize((224, 224)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor()
])
number_of_workers = 8
training_dataset = ImageFolder(root=os.path.join(dataset_directory,
"training"),
transform=data_transform)
training_dataset_loader = DataLoader(training_dataset,
batch_size=minibatch_size,
shuffle=True,
num_workers=number_of_workers)
validation_dataset = ImageFolder(root=os.path.join(dataset_directory,
"validation"),
transform=data_transform)
validation_dataset_loader = DataLoader(validation_dataset,
batch_size=minibatch_size,
shuffle=False,
num_workers=number_of_workers)
testing_dataset = ImageFolder(root=os.path.join(dataset_directory, "test"),
transform=data_transform)
testing_dataset_loader = DataLoader(testing_dataset,
batch_size=minibatch_size,
shuffle=False,
num_workers=number_of_workers)
return training_dataset_loader, validation_dataset_loader, testing_dataset_loader
def get_datasets(initial_pool):
transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(30),
transforms.ToTensor(),
transforms.Normalize(3 * [0.5], 3 * [0.5]),
])
test_transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(3 * [0.5], 3 * [0.5]),
])
# Note: We use the test set here as an example. You should make your own validation set.
train_ds = datasets.CIFAR10(".",
train=True,
transform=transform,
target_transform=None,
download=True)
test_set = datasets.CIFAR10(".",
train=False,
transform=test_transform,
target_transform=None,
download=True)
active_set = ActiveLearningDataset(
train_ds, pool_specifics={"transform": test_transform})
# We start labeling randomly.
active_set.label_randomly(initial_pool)
return active_set, test_set
def build_dataset(config):
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])
size = config['image_size']
train_transforms = transforms.Compose([
transforms.RandomResizedCrop(size=size, scale=(0.8, 1.0)),
transforms.RandomRotation(degrees=5),
transforms.ColorJitter(brightness=0.3, contrast=0.3, saturation=0.1),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize
])
val_transforms = transforms.Compose([
transforms.Resize(size=(size, size)),
transforms.ToTensor(),
normalize
])
dsTrain = ImageFolder('data/train/', train_transforms, target_transform=target_transform)
dsVal = ImageFolder('data/test/', val_transforms, target_transform=target_transform)
#dsTest = ImageFolder('data/WildFire/test/', val_transforms, target_transform=target_transform)
train_loader = DataLoader(dsTrain, batch_size=config['batch_size'], shuffle=True)
val_loader = DataLoader(dsVal, batch_size=config['batch_size'], shuffle=True)
#test_loader = DataLoader(dsTest, batch_size=config['batch_size'], shuffle=True)
return train_loader, val_loader
def __init__(self, args):
self.args = args
self.rvs = args.rvs
self.insp = args.insp
self.inch = args.inch
self.outf = args.outf
self.transformer = transforms.RandomChoice([
transforms.Compose([transforms.ToPILImage(), transforms.ColorJitter(brightness=0.1), transforms.ToTensor()]),
transforms.Compose([transforms.ToPILImage(), transforms.ColorJitter(contrast=0.1), transforms.ToTensor()]),
transforms.Compose([transforms.ToPILImage(), transforms.ColorJitter(saturation=0.1), transforms.ToTensor()]),
transforms.Compose([transforms.ToPILImage(), transforms.RandomRotation(degrees=10, resample=False, expand=False, center=None), transforms.ToTensor()]),
transforms.Compose([transforms.ToPILImage(), transforms.RandomHorizontalFlip(p=1), transforms.ToTensor()]),
])
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
self.D = D(insp=args.insp, inch=args.inch, outf=args.outf).to(self.device)
self.G = G(rvs=args.rvs).to(self.device)
self.criterion = nn.BCELoss()
self.optimizerD = torch.optim.Adam(self.D.parameters(), lr=args.lr, betas=(args.beta, 0.999))
self.optimizerG = torch.optim.Adam(self.G.parameters(), lr=args.lr, betas=(args.beta, 0.999))
self.test_noise = torch.randn(args.grid_w * args.grid_h, args.rvs, 1, 1).to(device=self.device)
self.real_label = 1
self.fake_label = 0
def __getitem__(self, idx):
img = self.load_image(idx)
img = transforms.RandomRotation(self.degree)(img)
root = self.label.loc[idx]['grapheme_root']
consonant = self.label.loc[idx]['consonant_diacritic']
vowel = self.label.loc[idx]['vowel_diacritic']
return transforms.ToTensor()(img), root, consonant, vowel
def __init__(self, hparams):
super().__init__()
self.data_dir = hparams["data_dir"] + "/snake_dataset/train"
self.batch_size = hparams.get("batch_size") or 64
self.train_val_split_ratio = hparams.get("train_val_split_ratio") or 0.9
self.num_workers = hparams.get("num_workers") or 1
self.pin_memory = hparams.get("pin_memory") or False
img_augmentation_transformations = [
# transforms.RandomAffine((-15, 15), translate=(0.2, 0.2)),
transforms.RandomHorizontalFlip(p=1.0),
transforms.RandomRotation((-15, 15)),
transforms.ColorJitter(hue=.05, saturation=.05),
]
self.transforms = transforms.Compose([
transforms.Resize((112, 112)),
transforms.ToTensor(),
transforms.RandomApply(img_augmentation_transformations, 0.5),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
self.data_train = None
self.data_val = None
def build_data_loader():
transform = transforms.Compose([
transforms.ColorJitter(brightness=0.1,
contrast=0.1,
saturation=0.1,
hue=0.1),
transforms.RandomRotation(90),
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
dataset = PlantSeedlingsDataset(os.environ['DATA_ROOT'],
transforms=transform)
train_set_size = len(dataset) * 7 // 10
val_set_size = len(dataset) - train_set_size
train_set, val_set = random_split(dataset, [train_set_size, val_set_size])
data_loader = {
'train':
DataLoader(train_set,
batch_size=BATCH_SIZE,
shuffle=True,
num_workers=NUM_WORKERS),
'val':
DataLoader(val_set,
batch_size=BATCH_SIZE,
shuffle=False,
num_workers=NUM_WORKERS),
}
return data_loader
def get_train_trans(SIZE):
return transforms.Compose([
transforms.CenterCrop(SIZE),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize(mean=[0, 0, 0], std=[0.3, 0.2, 0.1])
])
def hard_augment():
hard_augment_ = transforms.Compose([
transforms.RandomHorizontalFlip(p=0.5),
transforms.RandomRotation(degrees=(-90, 90)),
transforms.RandomVerticalFlip(p=0.5),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
])
return hard_augment_
def rotate_img(img: torch.Tensor) -> (torch.Tensor, float):
angle = torch.rand(1) * 360
rotate_transform = transforms.Compose([
transforms.ToPILImage(),
transforms.RandomRotation((angle, angle)),
transforms.ToTensor()
])
return rotate_transform(img), angle.item()
def __init__(self, resize_to=(224, 224)):
list_of_transforms = [
transforms.RandomVerticalFlip(),
transforms.RandomRotation(degrees=(-180, 180)),
transforms.Resize(size=resize_to),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
]
self.transform = transforms.Compose(list_of_transforms)
def __getitem__(self, idx):
img = self.load_image(idx)
img = transforms.RandomRotation(15)(img)
root = self.label.loc[idx]['grapheme_root']
consonant = self.label.loc[idx]['consonant_diacritic']
vowel = self.label.loc[idx]['vowel_diacritic']
timer = time.time()
unique = np.where(self.uniques == self.label.grapheme[idx])[0][0]
#print('unique: ', time.time() - timer)
return transforms.ToTensor()(img), root, consonant, vowel, unique
def get_train_trans(SIZE):
return transforms.Compose([
transforms.Resize((SIZE, SIZE)),
transforms.RandomHorizontalFlip(),
transforms.RandomVerticalFlip(),
transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
class Config:
data_root = args.path
training_size = 12500
test_size = 12500
lexpert = [[.55, .45], [.45, .55]]
if args.expertise == 0:
if args.case == 1:
expert_num = 10
elif args.case == 2:
expert_num = 25
elif args.case == 3:
expert_num = 12
as_expertise = np.array([
lexpert, lexpert, lexpert, lexpert, lexpert, lexpert, lexpert,
lexpert, lexpert, lexpert, lexpert, lexpert, lexpert, lexpert,
lexpert
])
elif args.expertise == 1:
if args.case == 1:
expert_num = 5
elif args.case == 2:
expert_num = 10
elif args.case == 3:
expert_num = 10
as_expertise = np.array([[[0.6, 0.4], [0.1, 0.9]],
[[0.8, 0.2], [0.4, 0.6]],
[[0.6, 0.4], [0.4, 0.6]],
[[0.7, 0.3], [0.3, 0.7]],
[[0.7, 0.3], [0.4, 0.6]]])
missing_label = np.array([0, 0, 0, 0, 0])
missing = False
num_classes = 2
left_input_size = 28 * 28
batch_size = 16
left_learning_rate = 1e-4
right_learning_rate = 1e-4
epoch_num = 20
device_id = args.device
experiment_case = args.case
train_transform = transforms.Compose([
transforms.Resize((150, 150), interpolation=2),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(45),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
test_transform = transforms.Compose([
transforms.Resize((150, 150), interpolation=2),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
def __init__(self, image_file, autoaugment=False):
super(ImageNetTrainingDataset, self).__init__()
self.image_file = image_file
# self.data = None
with open(self.image_file, "r") as file:
self.data = file.readlines()
# shuffle the dataset
for i in range(10):
random.shuffle(self.data)
self.imagenet_normalization_paramters = transforms.Normalize(
mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
# 先resize到512 再crop到448
# 用config来指定模型的size
self.model_size = ModelSize("resnet50_448")
model_size = self.model_size.imagesize_choice()
self.BASE_RESIZE_SIZE = model_size["resize"]
self.INPUT_SIZE = model_size["input"]
self.BRIGHTNESS = 0.4
self.HUE = 0.1
self.CONTRAST = 0.4
self.SATURATION = 0.4
# autoaugment
self.Autoaugment = autoaugment
self.index_sampler = [i for i in range(len(self.data))]
# 当前的数据增强【随机crop, 随机水平翻转，颜色变换，随机灰度，】
if self.Autoaugment:
self.image_transforms = transforms.Compose([
transforms.Resize(
(self.BASE_RESIZE_SIZE, self.BASE_RESIZE_SIZE),
Image.BILINEAR),
transforms.RandomCrop(self.INPUT_SIZE),
transforms.RandomHorizontalFlip(),
ImageNetPolicy(),
transforms.ToTensor(), self.imagenet_normalization_paramters
])
else:
self.image_transforms = transforms.Compose([
# transforms.RandomResizedCrop(self.INPUT_SIZE, scale=(0.2, 1.)),
transforms.Resize(
(self.BASE_RESIZE_SIZE, self.BASE_RESIZE_SIZE),
Image.BILINEAR),
transforms.RandomCrop(self.INPUT_SIZE),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(degrees=15),
transforms.ColorJitter(brightness=self.BRIGHTNESS,
contrast=self.CONTRAST,
hue=self.HUE,
saturation=self.SATURATION),
transforms.ToTensor(),
self.imagenet_normalization_paramters
])
def medium_augment():
medium_augment_ = transforms.Compose([
transforms.CenterCrop((100, 100)),
transforms.RandomCrop((80, 80)),
transforms.RandomHorizontalFlip(p=0.5),
transforms.RandomRotation(degrees=(-90, 90)),
transforms.RandomVerticalFlip(p=0.5),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
])
return medium_augment_
def preprocess_with_augmentation_without_norm(image_size):
return transforms.Compose([
transforms.Resize((image_size + 20, image_size + 20)),
transforms.RandomRotation(15, expand=True),
transforms.RandomCrop((image_size, image_size)),
transforms.RandomHorizontalFlip(),
transforms.ColorJitter(brightness=0.4,
contrast=0.4,
saturation=0.4,
hue=0.2),
transforms.ToTensor()
])
def gen_dataloaders(indir,
val_split=0.05,
shuffle=True,
batch_size=4,
seed=42,
img_size=224,
cuda=True):
data_transforms = {
'train':
transforms.Compose([
# transforms.ColorJitter(0.3, 0.3, 0.3, 0.3),
transforms.RandomResizedCrop(img_size, scale=(0.1, 1.0)),
transforms.RandomAffine(10.),
transforms.RandomRotation(13.),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
]),
'valid':
transforms.Compose([
transforms.Resize((img_size, img_size)),
transforms.ToTensor(),
])
}
mask_train_dataset = MaskDataset(path=indir,
transform=data_transforms['train'])
mask_valid_dataset = MaskDataset(path=indir,
transform=data_transforms['valid'])
# Creating data indices for training and validation splits:
dataset_size = len(mask_train_dataset)
indices = list(range(dataset_size))
split = int(np.floor(val_split * dataset_size))
if shuffle:
np.random.seed(seed)
np.random.shuffle(indices)
train_indices, val_indices = indices[split:], indices[:split]
# Creating data samplers and loaders
train_sampler = SubsetRandomSampler(train_indices)
valid_sampler = SubsetRandomSampler(val_indices)
kwargs = {'num_workers': 4, 'pin_memory': True} if cuda else {}
train_loader = torch.utils.data.DataLoader(mask_train_dataset,
batch_size=batch_size,
sampler=train_sampler,
**kwargs)
valid_loader = torch.utils.data.DataLoader(mask_valid_dataset,
batch_size=batch_size,
sampler=valid_sampler,
**kwargs)
return train_loader, valid_loader
def eyetransfrom(imsize=224):
eyetrans = transforms.Compose([
transforms.RandomHorizontalFlip(),
transforms.RandomVerticalFlip(),
transforms.RandomRotation(degrees=(0, 360)),
transforms.Resize(imsize),
transforms.CenterCrop(imsize),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
return eyetrans
def main():
args = args_parser()
utils.delete_path(args.log_dir)
utils.ensure_path(args.save_dir)
utils.ensure_path(args.log_dir)
utils.write_dict(vars(args), os.path.join(args.save_dir, 'arguments.csv'))
torch.manual_seed(args.seed)
cudnn.benchmark = True
torch.cuda.manual_seed_all(args.seed)
os.environ['CUDA_VISIBLE_DEVICES'] = args.device_ids
train_transformer = transforms.Compose([
transforms.Resize(size=(256, 256)),
transforms.RandomCrop(size=image_size),
transforms.RandomRotation(10),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
train_dataset = ZSLDataset(args.image_dir, args.train_file,
args.embed_file, args.attribute_file,
args.lable_name_file)
train_dataloader = train_dataset.get_dataloader(batch_size=args.batch_size,
num_workers=args.nw,
shuffle=True)
val_transformer = transforms.Compose([
transforms.Resize(size=image_size),
# transforms.CenterCrop(size=image_size),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
val_dataset = ZSLDataset(args.image_dir, args.val_file, args.embed_file,
args.attribute_file, args.lable_name_file)
val_dataloader = val_dataloader.get_dataloader(batch_size=args.batch_size,
num_workers=args.nw,
shuffle=False)
print(vars(args))
num_classes = train_dataset.get_class_size()
train_logger = SummaryLogger(args.log_dir)
trainer = FeatTrainer(num_classes,
backbone=args.backbone,
input_size=image_size,
train_logger=train_logger,
opt=args)
trainer.train(train_dataloader, val_dataloader)
def get_ood_dataloader(data_manager, batch_size: int = 16):
"""get_ood_dataloader [Returns OOD dataloader for Outlier exposure experiment]
Args:
data_manager ([object]): [data_manager]
batch_size (int, optional): [batch size for dataloader]. Defaults to 16.
Returns:
[torch.Dataloader]: [train loader]
[torch.Dataloader]: [ood lloader]
"""
train_X, train_y = data_manager.get_train_data()
ood_X, ood_y = data_manager.get_ood_data()
train_y = np.ones_like(train_y)
ood_y = np.zeros_like(ood_y)
train_X, train_y = train_X.astype(np.float32), train_y.astype(np.float32)
ood_X, ood_y = ood_X.astype(np.float32), ood_y.astype(np.float32)
train_X, train_y = torch.from_numpy(train_X), torch.from_numpy(train_y)
ood_X, ood_y = torch.from_numpy(ood_X), torch.from_numpy(ood_y)
train_dataset = DataHandler_For_Arrays(train_X, train_y)
transform_ood = transforms.Compose([
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
transforms.RandomCrop(size=32),
transforms.RandomRotation(degrees=(0, 180)),
transforms.RandomHorizontalFlip(),
])
outlier_data = DataHandler_For_Arrays(train_X,
train_y,
transform=transform_ood,
num_classes=1)
train_loader = DataLoader(
train_dataset,
sampler=RandomSampler(train_dataset),
batch_size=batch_size * 4,
num_workers=2,
pin_memory=pin_memory,
)
outlier_loader = DataLoader(
outlier_data,
sampler=RandomSampler(outlier_data),
batch_size=batch_size,
num_workers=2,
pin_memory=pin_memory,
)
return train_loader, outlier_loader
def generate_transform(transform_dict):
transform_list = []
for k, v in transform_dict.items():
if k.lower() == 'resize':
transform_list.append(transforms.Resize(v))
elif k.lower() == 'resize_nearest':
transform_list.append(
transforms.Resize(v, interpolation=Image.NEAREST))
elif k.lower() == 'crop':
transform_list.append(transforms.RandomCrop(v))
elif k.lower() == 'resized_crop':
assert len(v) == 4
h, w, min_scale = v
transform_list.append(
transforms.RandomResizedCrop((h, w), (min_scale, 1.)))
elif k.lower() == 'center_crop':
transform_list.append(transforms.CenterCrop(v))
elif k.lower() == 'center_resize':
transform_list.append(custom_transforms.CenterResize(v))
elif k.lower() == 'resize_center':
transform_list.append(
custom_transforms.CenterResize(v, resize_first=True))
elif k.lower() == 'h_flip':
transform_list.append(transforms.RandomHorizontalFlip())
elif k.lower() == 'rotate':
transform_list.append(transforms.RandomRotation(v))
elif k.lower() == 'color_jitter':
assert len(v) == 4
brightness, contrast, saturation, hue = v
transform_list.append(
transforms.ColorJitter(brightness, contrast, saturation, hue))
elif k.lower() == 'to_tensor':
transform_list.append(transforms.ToTensor())
elif k.lower() == 'normalize':
if v is None:
norm_value = ([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
elif len(v) == 6:
assert len(v) == 6
norm_value = ([v[0], v[1], v[2]], [v[3], v[4], v[5]])
elif len(v) == 2:
norm_value = ([v[0]], [v[1]])
else:
raise RuntimeError('Invalid arguments for normalize transform')
transform_list.append(
transforms.Normalize(mean=norm_value[0], std=norm_value[1]))
else:
print('Unsupported data transform type: {}'.format(k))
return transforms.Compose(transform_list)
def __init__(self, root, mode, batchsz, n_way, k_shot, k_query, resize, startidx=0):
"""
:param root: root path of mini-imagenet
:param mode: train, val or test
:param batchsz: batch size of sets, not batch of imgs
:param n_way:
:param k_shot:
:param k_query: num of qeruy imgs per class
:param resize: resize to
:param startidx: start to index label from startidx
"""
self.batchsz = batchsz # batch of set, not batch of imgs
self.n_way = n_way # n-way
self.k_shot = k_shot # k-shot
self.k_query = k_query # for evaluation
self.setsz = self.n_way * self.k_shot # num of samples per set
self.querysz = self.n_way * self.k_query # number of samples per set for evaluation
self.resize = resize # resize to
self.startidx = startidx # index label not from 0, but from startidx
print('%s, b:%d, %d-way, %d-shot, %d-query, resize:%d' % (mode, batchsz, n_way, k_shot, k_query, resize))
if mode == 'train':
self.transform = transforms.Compose([lambda x: Image.open(x).convert('RGB'),
transforms.RandomResizedCrop(self.resize, scale=(0.8, 1.0)),
transforms.RandomHorizontalFlip(),
# transforms.RandomVerticalFlip(),
transforms.RandomRotation(15),
transforms.ColorJitter(0.1, 0.1, 0.2, 0),
transforms.ToTensor(),
# transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
else:
self.transform = transforms.Compose([lambda x: Image.open(x).convert('RGB'),
transforms.Resize((self.resize, self.resize)),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
self.path = os.path.join(root, 'images') # image path
csvdata = self.loadCSV(os.path.join(root, mode + '.csv')) # csv path
self.data = []
self.img2label = {}
for i, (k, v) in enumerate(csvdata.items()):
self.data.append(v) # [[img1, img2, ...], [img111, ...]]
self.img2label[k] = i + self.startidx # {"img_name[:9]":label}
self.cls_num = len(self.data)
self.create_batch(self.batchsz)
def get_transformations(mean, std) -> Tuple[any, any]:
# normalization values for pretrained resnet on Imagenet
transform_train = transforms.Compose([
transforms.Resize((224, 224)),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(degrees=60),
transforms.ToTensor(),
transforms.Normalize(mean, std),
])
transform_test = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean, std),
])
return transform_train, transform_test
def __init__(self, data_dir, batch_size, labeled_percents=20):
self.mean = [0.485, 0.456, 0.406]
self.std = [0.229, 0.224, 0.225]
self.train_path = os.path.join(data_dir, "train")
self.classes = []
for d in os.listdir(self.train_path):
if ".txt" not in d:
self.classes.append(d)
self.transform = T.Compose([
T.Resize((128, 128), interpolation=2),
T.RandomRotation(45),
T.RandomVerticalFlip(),
T.RandomGrayscale(),
T.RandomSizedCrop((112, 112)),
T.ToTensor(),
T.Normalize(self.mean, self.std)
])
self.paths = []
self.true_labels = []
self.true_idxs = []
self.relabeled = []
self.labeled_idxs = []
self.unlabeled_idxs = []
ck = 0
last_cls = None
for i, c in enumerate(self.classes):
for file in list(
glob.glob(
os.path.join(self.train_path, os.path.join(c,
"*.*")))):
self.paths.append(file)
self.true_labels.append(c)
self.true_idxs.append(i)
self.relabeled.append(i)
unlabeled_limit = int(
(len(self.true_idxs) - ck) * labeled_percents)
for idx in range(ck, ck + unlabeled_limit):
self.relabeled[idx] = NO_LABEL
# collect indexes of labels both labeled and unlabeled
for i, l in enumerate(self.relabeled):
if l == -1:
self.unlabeled_idxs.append(i)
else:
self.labeled_idxs.append(i)