class RandomSizedCrop_64_512(BenchmarkTest):
def __init__(self):
self.augmentor_pipeline = Pipeline()
self.augmentor_pipeline.add_operation(
Operations.Crop(probability=1, width=64, height=64, centre=False))
self.augmentor_pipeline.add_operation(
Operations.Resize(probability=1,
width=512,
height=512,
resample_filter="BILINEAR"))
self.imgaug_transform = iaa.Sequential([
iaa.CropToFixedSize(width=64, height=64),
iaa.Scale(size=512, interpolation="linear")
])
self.solt_stream = slc.Stream([
slt.CropTransform(crop_size=(64, 64), crop_mode="r"),
slt.ResizeTransform(resize_to=(512, 512))
])
def albumentations(self, img):
img = albumentations.random_crop(img,
crop_height=64,
crop_width=64,
h_start=0,
w_start=0)
return albumentations.resize(img, height=512, width=512)
def augmentor(self, img):
for operation in self.augmentor_pipeline.operations:
img, = operation.perform_operation([img])
return np.array(img, np.uint8, copy=True)
def torchvision_transform(self, img):
img = torchvision.crop(img, top=0, left=0, height=64, width=64)
return torchvision.resize(img, (512, 512))
class BrightnessContrast(BenchmarkTest):
def __init__(self):
self.imgaug_transform = iaa.Sequential([
iaa.Multiply((1.5, 1.5), per_channel=False),
iaa.Add((127, 127), per_channel=False)
])
self.augmentor_pipeline = Pipeline()
self.augmentor_pipeline.add_operation(
Operations.RandomBrightness(probability=1,
min_factor=1.5,
max_factor=1.5))
self.augmentor_pipeline.add_operation(
Operations.RandomContrast(probability=1,
min_factor=1.5,
max_factor=1.5))
self.solt_stream = slc.Stream([
slt.ImageRandomBrightness(p=1, brightness_range=(127, 127)),
slt.ImageRandomContrast(p=1, contrast_range=(1.5, 1.5)),
])
def albumentations(self, img):
return albumentations.brightness_contrast_adjust(img,
alpha=1.5,
beta=0.5,
beta_by_max=True)
def torchvision_transform(self, img):
img = torchvision.adjust_brightness(img, brightness_factor=1.5)
img = torchvision.adjust_contrast(img, contrast_factor=1.5)
return img
def augmentor(self, img):
for operation in self.augmentor_pipeline.operations:
img, = operation.perform_operation([img])
return np.array(img, np.uint8, copy=True)
def create_aug_pipeline_train(input_size):
"""Image Augmentation Pipeline for Training Set."""
p_train = Pipeline()
# Random crop
p_train.add_operation(CropPercentageRange(probability=1, min_percentage_area=0.8, max_percentage_area=1, centre=False))
# Rotate the image by either 90, 180, or 270 degrees randomly
p_train.rotate_random_90(probability=0.5)
# Flip the image along its vertical axis
p_train.flip_top_bottom(probability=0.5)
# Flip the image along its horizontal axis
p_train.flip_left_right(probability=0.5)
# Random change brightness of the image
p_train.random_brightness(probability=0.5, min_factor=0.9, max_factor=1.1)
# Random change saturation of the image
p_train.random_color(probability=0.5, min_factor=0.9, max_factor=1.1)
# Resize the image to the desired input size of the model
p_train.resize(probability=1, width=input_size[0], height=input_size[1])
return p_train