def test_resize_in_memory():
tmpdir = tempfile.mkdtemp()
tmp = tempfile.NamedTemporaryFile(dir=tmpdir, suffix='.JPEG')
im = Image.new('RGB', original_dimensions)
im.save(tmp.name, 'JPEG')
resize = Operations.Resize(probability=1,
width=larger_dimensions[0],
height=larger_dimensions[1],
resample_filter="BICUBIC")
im = [im]
im_resized = resize.perform_operation(im)
assert im_resized[0].size == larger_dimensions
resize_smaller = Operations.Resize(probability=1,
width=smaller_dimensions[0],
height=smaller_dimensions[1],
resample_filter="BICUBIC")
im_resized_smaller = resize_smaller.perform_operation(im)
assert im_resized_smaller[0].size == smaller_dimensions
tmp.close()
def test_in_memory_distortions():
im, tmpdir = create_colour_temp_image((800, 800), "JPEG")
r_d = Operations.Distort(probability=1,
grid_width=8,
grid_height=8,
magnitude=8)
tmp_im = r_d.perform_operation([im])
assert tmp_im is not None
assert tmp_im[0].size == (800, 800)
im_bw, tmpdir_bw = create_greyscale_temp_image((800, 800), "PNG")
r_d_bw = Operations.Distort(probability=1,
grid_width=8,
grid_height=8,
magnitude=8)
tmp_im_bw = r_d_bw.perform_operation([im_bw])
assert tmp_im_bw is not None
assert tmp_im_bw[0].size == (800, 800)
assert isinstance(tmp_im_bw[0], Image.Image)
shutil.rmtree(tmpdir)
shutil.rmtree(tmpdir_bw)
def test_random_contrast_in_memory():
im, tmpdir = create_colour_temp_image((800, 800), "JPEG")
op = Operations.RandomContrast(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = op.perform_operation([im])
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
shutil.rmtree(tmpdir)
im_bw, tmpdir_bw = create_greyscale_temp_image((800, 800), "PNG")
op = Operations.RandomContrast(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = op.perform_operation([im_bw])
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
shutil.rmtree(tmpdir_bw)
def test_random_brightness_in_memory():
im, tmpdir = create_colour_temp_image((800, 800), "JPEG")
op = Operations.RandomBrightness(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = op.perform_operation([im])
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
shutil.rmtree(tmpdir)
if os.name == "nt":
# not testing bw images
return
im_bw, tmpdir_bw = create_greyscale_temp_image((800, 800), "PNG")
op = Operations.RandomBrightness(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = op.perform_operation([im_bw])
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
shutil.rmtree(tmpdir_bw)
def test_random_brightness_in_memory():
tmp, tmpdir = create_colour_temp_image((800, 800), "JPEG")
op = Operations.RandomBrightness(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = []
tmp_im.append(Image.open(tmp))
tmp_im = op.perform_operation(tmp_im)
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
tmp_bw, tmpdir_bw = create_greyscale_temp_image((800, 800), "PNG")
op = Operations.RandomBrightness(probability=1,
min_factor=0.5,
max_factor=1.5)
tmp_im = []
tmp_im.append(Image.open(tmp))
tmp_im = op.perform_operation(tmp_im)
assert tmp_im is not None
assert isinstance(tmp_im[0], Image.Image)
tmp.close()
tmp_bw.close()
shutil.rmtree(tmpdir)
shutil.rmtree(tmpdir_bw)
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 __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 __init__(self):
self.imgaug_transform = iaa.Add((127, 127), per_channel=False)
self.augmentor_op = Operations.RandomBrightness(probability=1,
min_factor=1.5,
max_factor=1.5)
self.solt_stream = slc.Stream(
[slt.ImageRandomBrightness(p=1, brightness_range=(127, 127))])
def __init__(self):
self.imgaug_transform = iaa.Multiply((1.5, 1.5), per_channel=False)
self.augmentor_op = Operations.RandomContrast(probability=1,
min_factor=1.5,
max_factor=1.5)
self.solt_stream = slc.Stream(
[slt.ImageRandomContrast(p=1, contrast_range=(1.5, 1.5))])
def __init__(self):
self.imgaug_transform = iaa.Scale(size=512, interpolation="linear")
self.solt_stream = slc.Stream([slt.Resize(resize_to=(512, 512))])
self.augmentor_op = Operations.Resize(probability=1,
width=512,
height=512,
resample_filter="BILINEAR")
def __init__(self):
self.imgaug_transform = iaa.CropToFixedSize(width=64, height=64)
self.augmentor_op = Operations.Crop(probability=1,
width=64,
height=64,
centre=False)
self.solt_stream = slc.Stream(
[slt.CropTransform(crop_size=(64, 64), crop_mode="r")])
def __init__(self):
self.imgaug_transform = iaa.Affine(rotate=(45, 45),
order=1,
mode="reflect")
self.augmentor_op = Operations.RotateStandard(probability=1,
max_left_rotation=45,
max_right_rotation=45)
self.solt_stream = slc.Stream(
[slt.RandomRotate(p=1, rotation_range=(45, 45))], padding="r")
def rotate_images(tmpdir, rot):
original_dimensions = (800, 800)
im_tmp = tmpdir.mkdir("subfolder").join('test.JPEG')
im = Image.new('RGB', original_dimensions)
im.save(str(im_tmp), 'JPEG')
r = Operations.Rotate(probability=1, rotation=rot)
im = [im]
im_r = r.perform_operation(im)
assert im_r is not None
assert im_r[0].size == original_dimensions
def test_rotate_images_custom_temp_files():
original_dimensions = (800, 800)
tmpdir = tempfile.mkdtemp()
tmp = create_temp_file(tmpdir, '.JPEG')
im = Image.new('RGB', original_dimensions)
im.save(tmp.name, 'JPEG')
r = Operations.Rotate(probability=1, rotation=90)
im = [im]
im_r = r.perform_operation(im)
assert im_r is not None
assert im_r[0].size == original_dimensions
shutil.rmtree(tmpdir)
def test_rotate_images_custom_temp_files():
original_dimensions = (800, 800)
tmpdir = tempfile.mkdtemp()
tmp = tempfile.NamedTemporaryFile(dir=tmpdir, suffix='.JPEG')
im = Image.new('RGB', original_dimensions)
im.save(tmp.name, 'JPEG')
r = Operations.Rotate(probability=1, rotation=90)
im_r = r.perform_operation(im)
assert im_r is not None
assert im_r.size == original_dimensions
tmp.close()
shutil.rmtree(tmpdir)
def test_zoom_min_max():
original_dimensions = (800, 800)
tmpdir = tempfile.mkdtemp()
tmp = None
try:
tmp = tempfile.NamedTemporaryFile(dir=tmpdir, suffix='.JPEG')
im = Image.new('RGB', original_dimensions)
im.save(tmp.name, 'JPEG')
r = Operations.ZoomRandom(probability=1, min_zoom=0.5, max_zoom=0.9)
im_r = r.perform_operation(im)
assert im_r is not None
assert im_r.size == original_dimensions
finally:
if tmp is not None:
tmp.close()
shutil.rmtree(tmpdir)
def __init__(self):
self.augmentor_op = Operations.Greyscale(probability=1)
self.imgaug_transform = iaa.Grayscale(alpha=1.0)
self.solt_stream = slc.Stream([slt.ImageColorTransform(mode="rgb2gs")])
def get_augmentation_group(data_aug_group,
input_size,
center=True,
resize=True):
resize_prob = 1 if resize else 0
center_prob = 1 if center else 0
DATA_AUGMENTATION_GROUPS = [
# GROUP 0 (NO DATA AUGMENTATION)
[
# Center crop
CropCenter(probability=center_prob),
# Resize the image to the desired input size
Operations.Resize(probability=resize_prob,
width=input_size[0],
height=input_size[1],
resample_filter="BICUBIC")
],
# GROUP 1 (Common transformations: rotations, flips, crops, shears)
[
# Rotate the image by 90 degrees randomly
Operations.Rotate(probability=0.5, rotation=-1),
# Flip top/bottom
Operations.Flip(probability=0.5,
top_bottom_left_right="TOP_BOTTOM"),
# Flip left/right
Operations.Flip(probability=0.5,
top_bottom_left_right="LEFT_RIGHT"),
# Random crops
Operations.CropPercentage(probability=0.5,
percentage_area=0.85,
centre=center,
randomise_percentage_area=True),
# Resize the image to the desired input size
Operations.Resize(probability=resize_prob,
width=input_size[0],
height=input_size[1],
resample_filter="BICUBIC")
],
# GROUP 2 (Pixel intensity transformations)
[
# Rotate the image by 90 degrees randomly
Operations.Rotate(probability=0.5, rotation=-1),
# Flip top/bottom
Operations.Flip(probability=0.5,
top_bottom_left_right="TOP_BOTTOM"),
# Flip left/right
Operations.Flip(probability=0.5,
top_bottom_left_right="LEFT_RIGHT"),
# Random change brightness of the image
Operations.RandomBrightness(probability=0.5,
min_factor=0.5,
max_factor=1.5),
# Random change saturation of the image
Operations.RandomColor(probability=0.5,
min_factor=0.5,
max_factor=1.5),
# Random change saturation of the image
Operations.RandomContrast(probability=0.5,
min_factor=0.5,
max_factor=1.5),
# Random crops
Operations.CropPercentage(probability=0.5,
percentage_area=0.85,
centre=center,
randomise_percentage_area=True),
# Resize the image to the desired input size
Operations.Resize(probability=resize_prob,
width=input_size[0],
height=input_size[1],
resample_filter="BICUBIC")
],
# GROUP 3 (Perspective transformations)
[
# Rotate the image by 90 degrees randomly
Operations.Rotate(probability=0.5, rotation=-1),
# Flip top/bottom
Operations.Flip(probability=0.5,
top_bottom_left_right="TOP_BOTTOM"),
# Flip left/right
Operations.Flip(probability=0.5,
top_bottom_left_right="LEFT_RIGHT"),
# Shear Image
Operations.Shear(probability=0.5,
max_shear_left=20,
max_shear_right=20),
# Random Tilt up down
Operations.Skew(probability=0.5,
skew_type="TILT_TOP_BOTTOM",
magnitude=1.0),
# Random Tilt left right
Operations.Skew(probability=0.5,
skew_type="TILT_LEFT_RIGHT",
magnitude=1.0),
# Random Skew CORNER
Operations.Skew(probability=0.5, skew_type="CORNER",
magnitude=1.3),
# Random crops
Operations.CropPercentage(probability=0.5,
percentage_area=0.85,
centre=center,
randomise_percentage_area=True),
# Resize the image to the desired input size
Operations.Resize(probability=resize_prob,
width=input_size[0],
height=input_size[1],
resample_filter="BICUBIC")
],
# GROUP 4 (Noise transformations)
[
# Center crop
CropCenter(probability=1),
# Rotate the image by 90 degrees randomly
Operations.Rotate(probability=0.5, rotation=-1),
# Flip top/bottom
Operations.Flip(probability=0.5,
top_bottom_left_right="TOP_BOTTOM"),
# Flip left/right
Operations.Flip(probability=0.5,
top_bottom_left_right="LEFT_RIGHT"),
# Random distortions
Operations.Distort(probability=0.5,
grid_width=8,
grid_height=8,
magnitude=15),
# Random erasing
Operations.RandomErasing(probability=0.5, rectangle_area=0.25),
# Random crops
Operations.CropPercentage(probability=0.5,
percentage_area=0.85,
centre=center,
randomise_percentage_area=True),
# Resize the image to the desired input size
Operations.Resize(probability=resize_prob,
width=input_size[0],
height=input_size[1],
resample_filter="BICUBIC")
]
]
return DATA_AUGMENTATION_GROUPS[data_aug_group]
def __init__(self):
self.imgaug_transform = iaa.AllChannelsHistogramEqualization()
self.augmentor_op = Operations.HistogramEqualisation(probability=1)
def __init__(self):
self.imgaug_transform = iaa.Fliplr(p=1)
self.augmentor_op = Operations.Flip(probability=1,
top_bottom_left_right="LEFT_RIGHT")
self.solt_stream = slc.Stream([slt.Flip(p=1, axis=1)])
def __init__(self):
self.imgaug_transform = iaa.Flipud(p=1)
self.augmentor_op = Operations.Flip(probability=1,
top_bottom_left_right="TOP_BOTTOM")
self.solt_stream = slc.Stream([slt.RandomFlip(p=1, axis=0)])
