def __im_preproc(self, image: np.ndarray) -> np.ndarray:
image = cv2.resize(image, tuple(self.im_shape))
blur = Blur(p=0.9, blur_limit=7)
hsv = HueSaturationValue(p=0.9)
crop = RandomCrop(p=1, height=self.im_shape[1], width=self.im_shape[0])
image = blur.apply(image, **blur.get_params())
image = hsv.apply(image, **hsv.get_params())
image = crop.apply(image, **crop.get_params())
if self.grayscale:
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
return image[:, :, np.newaxis]
def get_data_transforms(self):
cfg = self.cfg
bbox_params = self.get_bbox_params()
transform = Compose([Resize(cfg.data.img_sz, cfg.data.img_sz)],
bbox_params=bbox_params)
augmentors_dict = {
'Blur': Blur(),
'RandomRotate90': RandomRotate90(),
'HorizontalFlip': HorizontalFlip(),
'VerticalFlip': VerticalFlip(),
'GaussianBlur': GaussianBlur(),
'GaussNoise': GaussNoise(),
'RGBShift': RGBShift(),
'ToGray': ToGray()
}
aug_transforms = []
for augmentor in cfg.data.augmentors:
try:
aug_transforms.append(augmentors_dict[augmentor])
except KeyError as e:
log.warning(
'{0} is an unknown augmentor. Continuing without {0}. \
Known augmentors are: {1}'.format(
e, list(augmentors_dict.keys())))
aug_transforms.append(Resize(cfg.data.img_sz, cfg.data.img_sz))
aug_transform = Compose(aug_transforms, bbox_params=bbox_params)
return transform, aug_transform
def get_data_transforms(self) -> Tuple[BasicTransform, BasicTransform]:
"""Get albumentations transform objects for data augmentation.
Returns:
1st tuple arg: a transform that doesn't do any data augmentation
2nd tuple arg: a transform with data augmentation
"""
cfg = self.cfg
bbox_params = self.get_bbox_params()
transform = Compose([Resize(cfg.data.img_sz, cfg.data.img_sz)],
bbox_params=bbox_params)
augmentors_dict = {
'Blur': Blur(),
'RandomRotate90': RandomRotate90(),
'HorizontalFlip': HorizontalFlip(),
'VerticalFlip': VerticalFlip(),
'GaussianBlur': GaussianBlur(),
'GaussNoise': GaussNoise(),
'RGBShift': RGBShift(),
'ToGray': ToGray()
}
aug_transforms = []
for augmentor in cfg.data.augmentors:
try:
aug_transforms.append(augmentors_dict[augmentor])
except KeyError as e:
log.warning(
'{0} is an unknown augmentor. Continuing without {0}. \
Known augmentors are: {1}'.format(
e, list(augmentors_dict.keys())))
aug_transforms.append(Resize(cfg.data.img_sz, cfg.data.img_sz))
aug_transform = Compose(aug_transforms, bbox_params=bbox_params)
return transform, aug_transform
def oneof_always_apply_crash():
aug = Compose(
[HorizontalFlip(),
Rotate(),
OneOf([Blur(), MedianBlur()], p=1)], p=1)
image = np.ones((8, 8))
data = aug(image=image)
assert data
def test_deterministic_one_or_other():
aug = ReplayCompose([OneOrOther(HorizontalFlip(), Blur())], p=1)
for i in range(10):
image = (np.random.random((8, 8)) * 255).astype(np.uint8)
image2 = np.copy(image)
data = aug(image=image)
assert "replay" in data
data2 = ReplayCompose.replay(data["replay"], image=image2)
assert np.array_equal(data["image"], data2["image"])
def build_databunch(data_dir, img_sz, batch_sz, class_names, augmentors):
num_workers = 4
train_dir = join(data_dir, 'train')
valid_dir = join(data_dir, 'valid')
augmentors_dict = {
'Blur': Blur(),
'RandomRotate90': RandomRotate90(),
'HorizontalFlip': HorizontalFlip(),
'VerticalFlip': VerticalFlip(),
'GaussianBlur': GaussianBlur(),
'GaussNoise': GaussNoise(),
'RGBShift': RGBShift(),
'ToGray': ToGray()
}
aug_transforms = []
for augmentor in augmentors:
try:
aug_transforms.append(augmentors_dict[augmentor])
except KeyError as e:
log.warning('{0} is an unknown augmentor. Continuing without {0}. \
Known augmentors are: {1}'.format(e,
list(
augmentors_dict.keys())))
aug_transforms = Compose(aug_transforms)
train_ds = AlbumentationDataset(ImageFolder(train_dir,
classes=class_names),
transform=aug_transforms)
valid_ds = AlbumentationDataset(ImageFolder(valid_dir,
classes=class_names))
train_dl = DataLoader(train_ds,
shuffle=True,
batch_size=batch_sz,
num_workers=num_workers,
drop_last=True,
pin_memory=True)
valid_dl = DataLoader(valid_ds,
batch_size=batch_sz,
num_workers=num_workers,
pin_memory=True)
return DataBunch(train_ds, train_dl, valid_ds, valid_dl, class_names)
def get_transform():
transforms = A.Compose(
[
#HorizontalFlip(p=0.5),
#VerticalFlip(p=0.5),
#RandomSizedBBoxSafeCrop(700, 700, erosion_rate=0.0, interpolation=1, always_apply=False, p=0.5),
Blur(blur_limit=7, always_apply=False, p=0.5),
RandomBrightness(limit=0.2, always_apply=False, p=0.5),
#Downscale(scale_min=0.5, scale_max=0.9, interpolation=0, always_apply=False, p=0.5),
#PadIfNeeded(min_height=1024, min_width=1024, pad_height_divisor=None, pad_width_divisor=None, border_mode=4, value=None, mask_value=None, always_apply=False, p=1.0),
#RandomFog(fog_coef_lower=0.3, fog_coef_upper=1, alpha_coef=0.08, always_apply=False, p=0.2),
#RandomRain(slant_lower=-10, slant_upper=10, drop_length=20, drop_width=1, drop_color=(200, 200, 200), p=0.2),
#RandomSnow(snow_point_lower=0.1, snow_point_upper=0.3, brightness_coeff=2.5, always_apply=False, p=0.2)
],
bbox_params=A.BboxParams(format='pascal_voc',
label_fields=['bbox_classes']))
return transforms
def test_per_channel_multi():
transforms = [Blur(), Rotate()]
augmentation = PerChannel(transforms, p=1)
image = np.ones((8, 8, 5))
data = augmentation(image=image)
assert data
def build_databunch(data_dir, img_sz, batch_sz, class_names, rare_classes,
desired_prob, augmentors):
num_workers = 4
train_dir = join(data_dir, 'train')
valid_dir = join(data_dir, 'valid')
augmentors_dict = {
'Blur': Blur(),
'RandomRotate90': RandomRotate90(),
'HorizontalFlip': HorizontalFlip(),
'VerticalFlip': VerticalFlip(),
'GaussianBlur': GaussianBlur(),
'GaussNoise': GaussNoise(),
'RGBShift': RGBShift(),
'ToGray': ToGray()
}
aug_transforms = []
for augmentor in augmentors:
try:
aug_transforms.append(augmentors_dict[augmentor])
except KeyError as e:
log.warning('{0} is an unknown augmentor. Continuing without {0}. \
Known augmentors are: {1}'.format(e,
list(
augmentors_dict.keys())))
aug_transforms = Compose(aug_transforms)
train_ds = AlbumentationDataset(ImageFolder(train_dir,
classes=class_names),
transform=aug_transforms)
valid_ds = AlbumentationDataset(ImageFolder(valid_dir,
classes=class_names))
if rare_classes != []:
targets = [target for _, target in train_ds.orig_dataset.imgs]
train_sample_weights = calculate_oversampling_weights(
targets, rare_classes, desired_prob)
num_train_samples = len(train_ds)
train_sampler = WeightedRandomSampler(weights=train_sample_weights,
num_samples=num_train_samples,
replacement=True)
shuffle = False
else:
train_sampler = None
shuffle = True
train_dl = DataLoader(train_ds,
shuffle=shuffle,
batch_size=batch_sz,
num_workers=num_workers,
drop_last=True,
pin_memory=True,
sampler=train_sampler)
valid_dl = DataLoader(valid_ds,
batch_size=batch_sz,
num_workers=num_workers,
pin_memory=True)
return DataBunch(train_ds, train_dl, valid_ds, valid_dl, class_names)