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_random_rotate():
image = np.ones((192, 192, 3))
bboxes = [[78, 42, 142, 80]]
aug = Rotate(limit=15, p=1.)
transformed = aug(image=image, bboxes=bboxes)
assert len(bboxes) == len(transformed['bboxes'])
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
0.406], [0.229, 0.224, 0.225, 0.229, 0.224, 0.225])
ratio = 273 * 6 / 256
w, h = sz, int(ratio * sz)
n_epochs = int(params['n_epochs'])
TTA = int(params['TTA'])
balanced_sampler = bool(int(params['balanced_sampler']))
if balanced_sampler:
sampler = BalanceClassSampler
else:
sampler = None
train_aug = Compose(
[
OneOf([], p=0.20),
# HorizontalFlip(0.4),
VerticalFlip(0.4),
Rotate(limit=360, border_mode=2, p=0.4),
Resize(sz, sz, p=1, always_apply=True),
RandomSizedCrop(min_max_height=(int(sz * 0.8), int(sz * 0.8)),
height=sz,
width=sz,
p=0.4),
Resize(sz, sz, p=1, always_apply=True)
], )
val_aug = Compose([Resize(sz, sz, p=1, always_apply=True)])
# val_aug = None
data_dir = params['data_dir']
image_path = params['image_path']
test_image_path = params['test_image_path']