def make_voc_transforms(image_set, remove_difficult):
normalize = T.Compose([
T.ToTensor(),
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
transform_train = T.Compose([
T.RandomHorizontalFlip(),
T.RandomSelect(
T.RandomResize([400, 500, 600], max_size=1000),
T.Compose([
T.RandomResize([400, 500, 600]),
T.RandomCrop((384, 384)),
T.RandomResize([400, 500, 600], max_size=1000),
]),
),
normalize,
T.RemoveDifficult(remove_difficult),
])
transform_val = T.Compose(
[T.RandomResize([600], max_size=1000), normalize])
transforms = {
"train": transform_train,
"trainval": transform_train,
"val": transform_val,
"test": transform_val,
}
return transforms[image_set]
def make_lvis_transforms(image_set):
normalize = T.Compose([
T.ToTensor(),
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
transform_train = T.Compose([
T.RandomHorizontalFlip(),
T.RandomSelect(
T.RandomResize(scales, max_size=1333),
T.Compose([
T.RandomResize([400, 500, 600]),
T.RandomCrop((384, 384)),
T.RandomResize(scales, max_size=1333),
]),
),
normalize,
])
transform_val = T.Compose(
[T.RandomResize([800], max_size=1333), normalize])
transforms = {
"train": transform_train,
"trainval": transform_train,
"val": transform_val,
"test": transform_val,
}
return transforms[image_set]
def get_transform(train, resolution):
transforms = []
# if square resolution, perform some aspect cropping
# otherwise, resize to the resolution as specified
if resolution[0] == resolution[1]:
base_size = resolution[0] + 32
crop_size = resolution[0]
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms.append(T.RandomResize(min_size, max_size))
# during training mode, perform some data randomization
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomCrop(crop_size))
else:
# transforms.append(T.Resize(resolution))
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.ToTensor())
transforms.append(
T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def get_transform(train):
base_size = 520
crop_size = 480
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms = []
transforms.append(T.RandomResize(min_size, max_size))
if train:
transforms.append(
T.RandomColorJitter(brightness=0.25,
contrast=0.25,
saturation=0.25,
hue=0.25))
transforms.append(T.RandomGaussianSmoothing(radius=[0, 5]))
transforms.append(T.RandomRotation(degrees=30, fill=0))
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomPerspective(fill=0))
transforms.append(T.RandomCrop(crop_size, fill=0))
transforms.append(T.RandomGrayscale(p=0.1))
transforms.append(T.ToTensor())
transforms.append(
T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def __init__(self,
base_size,
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)):
self.transforms = T.Compose([
T.RandomResize(base_size, base_size),
T.ToTensor(),
T.Normalize(mean=mean, std=std),
])
def get_transform(train):
base_size = crop_size = 385
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms = []
transforms.append(T.RandomResize(min_size, max_size))
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.ToTensor())
return T.Compose(transforms)
def __init__(self,
base_size,
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225)):
self.transforms = T.Compose([
T.RandomResize(base_size, base_size),
T.PILToTensor(),
T.ConvertImageDtype(torch.float),
T.Normalize(mean=mean, std=std),
])
def __init__(self, base_size, crop_size, hflip_prob=0.5, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)):
min_size = int(0.5 * base_size)
max_size = int(2.0 * base_size)
trans = [T.RandomResize(min_size, max_size)]
if hflip_prob > 0:
trans.append(T.RandomHorizontalFlip(hflip_prob))
trans.extend([
T.RandomCrop(crop_size),
T.ToTensor(),
T.Normalize(mean=mean, std=std),
])
self.transforms = T.Compose(trans)
def __init__(self):
# Load all the images from coco
self.images = glob.glob('../data/coco/*/*.jpg')
# self data size
self.data_size = len(self.images)
# define the required transform on images
self.resize = T.RandomResize([800], max_size=1331)
self.transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
def get_transforms(train, base_size=520, crop_size=480):
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms = []
transforms.append(T.RandomResize(min_size, max_size))
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomCrop(crop_size))
transforms.append(T.ToTensor())
transforms.append(T.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def get_transform(mode, base_size):
#base_size = 520
#crop_size = 480
crop_size = int(480 * base_size / 520)
min_size = int((0.5 if mode == 'train' else 1.0) * base_size)
max_size = int((2.0 if mode == 'train' else 1.0) * base_size)
transforms = []
transforms.append(T.RandomResize(min_size, max_size))
if mode == 'train':
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomCrop(crop_size))
transforms.append(T.ToTensor())
transforms.append(
T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def __init__(self,
base_size,
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225),
contrast=1,
brightness=1,
sigma=1):
self.contrast_initial = contrast
self.contrast_final = contrast
if (contrast == 1):
self.contrast_final = contrast
else:
self.contrast_final = self.contrast_initial - 1
self.brightness_initial = brightness
self.brightness_final = brightness
if (brightness == 1):
self.brightness_final = brightness
else:
self.brightness_final = self.brightness_initial - 1
self.sigma_initial = sigma
self.sigma_final = sigma
if (sigma == 1):
self.sigma_final = sigma
else:
self.sigma_final = self.sigma_initial - 1
print("Contrast: ({}, {})".format(self.contrast_final,
self.contrast_initial))
print("Brightness: ({}, {})".format(self.brightness_final,
self.brightness_initial))
print("Sigma: ({}, {})".format(self.sigma_final, self.sigma_initial))
self.transforms = T.Compose([
T.RandomResize(base_size, base_size),
T.PILToTensor(),
T.ConvertImageDtype(torch.float),
T.ColorJitter(contrast=(self.contrast_final,
self.contrast_initial),
brightness=(self.brightness_final,
self.brightness_initial)),
#T.GaussianBlur(kernel_size=19, sigma=(self.sigma_final, self.sigma_initial)),
T.Normalize(mean=mean, std=std),
])
def get_transform(train):
base_size = 1000
crop_size = 768
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms = []
transforms.append(T.RandomResize(min_size, max_size))
if train:
transforms.append(T.ColorJitter(0.5, 0.5, 0.5, 0.5))
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomVerticalFlip(0.5))
transforms.append(T.RandomCrop(crop_size))
transforms.append(T.ToTensor())
transforms.append(
T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def get_transform(train, resolution):
transforms = []
# if square resolution, perform some aspect cropping
# otherwise, resize to the resolution as specified
if resolution[0] == resolution[1]:
base_size = resolution[0] + 32 #520
crop_size = resolution[0] #480
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms.append(T.RandomResize(min_size, max_size))
# during training mode, perform some data randomization
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.RandomCrop(crop_size))
if (args.randomize_color > 0.0):
transforms.append(T.RandomizeColor(args.randomize_color))
else:
transforms.append(T.Resize(resolution))
if train:
transforms.append(T.RandomHorizontalFlip(0.5))
if (args.randomize_color > 0.0):
transforms.append(T.RandomizeColor(args.randomize_color))
transforms.append(
T.IncreaseContrastSaturation(
args.adjust_contrast_saturation_val))
transforms.append(T.ToTensor())
mean_norm = args.mean_norm
std_norm = args.std_norm
transforms.append(
T.Normalize(mean=[mean_norm, mean_norm, mean_norm],
std=[std_norm, std_norm, std_norm]))
return T.Compose(transforms)
def get_transform(train):
base_size = 520
crop_size = 480
min_size = int((0.5 if train else 1.0) * base_size)
max_size = int((2.0 if train else 1.0) * base_size)
transforms = []
if train:
transforms.append(T.RandomResize(
min_size, max_size)) #move this line inside if condition
transforms.append(T.RandomHorizontalFlip(0.5))
# can add random rotation if have chance
transforms.append(T.RandomCrop(crop_size))
else:
transforms.append(T.Resize(
(min_size, max_size)
)) # add new line to reshape test/dp-train data into square shape
transforms.append(T.ToTensor())
transforms.append(
T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
return T.Compose(transforms)
def build_transforms(is_train, size, crop_size,mode="baseline"):
mean = (0.485, 0.456, 0.406)
std = (0.229, 0.224, 0.225)
fill = tuple([int(v * 255) for v in mean])
ignore_value = 255
transforms=[]
min_scale=1
max_scale=1
if is_train:
min_scale=0.5
max_scale=2
transforms.append(T.RandomResize(int(min_scale*size),int(max_scale*size)))
if is_train:
if mode=="baseline":
pass
elif mode=="randaug":
transforms.append(T.RandAugment(2,1/3,prob=1.0,fill=fill,ignore_value=ignore_value))
elif mode=="custom1":
transforms.append(T.ColorJitter(0.5,0.5,(0.5,2),0.05))
transforms.append(T.AddNoise(10))
transforms.append(T.RandomRotation((-10,10), mean=fill, ignore_value=0))
else:
raise NotImplementedError()
transforms.append(
T.RandomCrop(
crop_size,crop_size,
fill,
ignore_value,
random_pad=is_train
))
transforms.append(T.RandomHorizontalFlip(0.5))
transforms.append(T.ToTensor())
transforms.append(T.Normalize(
mean,
std
))
return T.Compose(transforms)
