def someAug():
bg = iap.Uniform(16, 18)
#Creating a series of augmentations
shearXY = iaa.Sequential([
iaa.ShearY(iap.Uniform(-10, 10), cval=bg),
iaa.ShearX(iap.Uniform(-10, 10), cval=bg)
])
rotate = iaa.Rotate(rotate=iap.Choice([-30, -15, 15, 30],
p=[0.25, 0.25, 0.25, 0.25]),
cval=bg)
pwAff = iaa.PiecewiseAffine(scale=(0.01, 0.06), cval=bg)
affine = iaa.Affine(scale={
"x": iap.Uniform(1.1, 1.2),
"y": iap.Uniform(1.1, 1.2)
},
cval=bg)
noise = iaa.AdditiveGaussianNoise(loc=0, scale=(0, 0.025 * 255))
#Using SomeOf to randomly select some augmentations
someAug = iaa.SomeOf(iap.Choice([2, 3, 4], p=[1 / 3, 1 / 3, 1 / 3]),
[affine, shearXY, pwAff, rotate, noise],
random_order=True)
return someAug
def _lane_argue(*, image, lane_src):
lines_tuple = [[(float(pt['x']), float(pt['y'])) for pt in line_spec] for line_spec in lane_src['Lines']]
lss = [ia_LineString(line_tuple_spec) for line_tuple_spec in lines_tuple]
lsoi = LineStringsOnImage(lss, shape=image.shape)
color_shift = iaa.OneOf([
iaa.GaussianBlur(sigma=(0.5, 1.5)),
iaa.LinearContrast((1.5, 1.5), per_channel=False),
iaa.Multiply((0.8, 1.2), per_channel=0.2),
iaa.AdditiveGaussianNoise(loc=0, scale=(0.0, 0.1 * 255), per_channel=0.5),
iaa.WithColorspace(to_colorspace=iaa.CSPACE_HSV, from_colorspace=iaa.CSPACE_RGB,
children=iaa.WithChannels(0, iaa.Multiply((0.7, 1.3)))),
iaa.WithColorspace(to_colorspace=iaa.CSPACE_HSV, from_colorspace=iaa.CSPACE_RGB,
children=iaa.WithChannels(1, iaa.Multiply((0.1, 2)))),
iaa.WithColorspace(to_colorspace=iaa.CSPACE_HSV, from_colorspace=iaa.CSPACE_RGB,
children=iaa.WithChannels(2, iaa.Multiply((0.5, 1.5)))),
])
posion_shift = iaa.SomeOf(4, [
iaa.Fliplr(),
iaa.Crop(percent=([0, 0.2], [0, 0.15], [0, 0], [0, 0.15]), keep_size=True),
iaa.TranslateX(px=(-16, 16)),
iaa.ShearX(shear=(-15, 15)),
iaa.Rotate(rotate=(-15, 15))
])
aug = iaa.Sequential([
iaa.Sometimes(p=0.6, then_list=color_shift),
iaa.Sometimes(p=0.6, then_list=posion_shift)
], random_order=True)
batch = ia.Batch(images=[image], line_strings=[lsoi])
batch_aug = list(aug.augment_batches([batch]))[0] # augment_batches returns a generator
image_aug = batch_aug.images_aug[0]
lsoi_aug = batch_aug.line_strings_aug[0]
lane_aug = [[dict(x=kpt.x, y=kpt.y) for kpt in shapely_line.to_keypoints()] for shapely_line in lsoi_aug]
return image_aug, dict(Lines=lane_aug)
def rotation(degree, input_path, output_path, image_count):
images = []
labels = []
for img_path in range(image_count):
img = imageio.imread(input_path + '/images/' + str(img_path) + '.png')
images.append(img)
lbl = imageio.imread(input_path + '/labels/' + str(img_path) + '.png')
labels.append(lbl)
seq = iaa.Sequential(
[
iaa.Rotate((degree)),
]
)
images_aug = seq(images=images)
labels_aug = seq(images=labels)
path = os.path.join(output_path, 'images')
os.mkdir(path)
path = os.path.join(output_path, 'labels')
os.mkdir(path)
for indx, i in enumerate(images_aug):
imageio.imwrite(output_path + '/images/' + 'rotat'+ '_' + str(indx) + '.png', i)
for indx, i in enumerate(labels_aug):
imageio.imwrite(output_path + '/labels/' + 'rotat'+ '_' + str(indx) + '.png', i)
print("Rotation results were saved given directory.")
def augmentation(X, y, masks, itterations):
ys = np.concatenate(
(y[:, :, np.newaxis], y[:, :, np.newaxis], y[:, :, np.newaxis]),
axis=2)
masks = np.concatenate(
(masks[:, :, np.newaxis], masks[:, :, np.newaxis], masks[:, :,
np.newaxis]),
axis=2)
B = np.concatenate((ys, X, masks), axis=2).astype(np.uint8)
A = np.zeros((itterations, np.shape(B)[0], np.shape(B)[1], np.shape(B)[2]))
for num in range(itterations):
aug1 = iaa.Affine(scale={"x": (1, 1.5), "y": (1, 1.5)})
aug2 = iaa.Fliplr(1)
aug3 = iaa.Flipud(1)
aug4 = iaa.ShearX((-20, 20), mode='reflect')
aug5 = iaa.ShearY((-20, 20), mode='reflect')
aug6 = iaa.Rotate((-45, 45), mode='reflect')
aug7 = iaa.WithChannels((3, 4, 5), iaa.Multiply((0.5, 1.5)))
aug_list = [aug1, aug2, aug3, aug4, aug5, aug6, aug7]
ID = np.random.randint(0, len(aug_list))
A[num, :, :, :] = aug_list[ID](image=B)
aug_y = A[:, :, :, 0]
aug_X = A[:, :, :, 3:6]
aug_masks = A[:, :, :, 8]
return aug_X, aug_y, aug_masks
def __init__(self):
"""
init Constructor
@param self Object.
"""
self.aug = iaa.Sequential([
# Blur or Sharpness
iaa.Sometimes(
0.25,
iaa.OneOf([
iaa.GaussianBlur(sigma=(0, 1.0)),
iaa.pillike.EnhanceSharpness(factor=(0.8, 1.5))
])),
# Flip horizontally
iaa.Fliplr(0.5),
# Rotation
iaa.Rotate((-20, 20)),
# Pixel Dropout
iaa.Sometimes(
0.25,
iaa.OneOf([
iaa.Dropout(p=(0, 0.1)),
iaa.CoarseDropout(0.1, size_percent=0.5)
])),
# Color
iaa.AddToHueAndSaturation(value=(-10, 10), per_channel=True),
])
def augment_3(self, image, depth, lidar, crop_size, degree):
# rsz = iaa.Resize({"height": resize_size[0], "width": resize_size[1]})
seq = iaa.Sequential(
[
iaa.PadToFixedSize(height=crop_size[0],
width=crop_size[1]), # 保证可crop
iaa.CropToFixedSize(height=crop_size[0],
width=crop_size[1]), # random crop
iaa.Fliplr(0.5),
# iaa.Flipud(0.5),
iaa.Rotate((-degree, degree)),
iaa.GammaContrast((0.9, 1.1)),
iaa.Multiply((0.9, 1.1)),
],
random_order=True)
depth, lidar = np.expand_dims(depth, 2), np.expand_dims(lidar, 2)
tmp = np.concatenate((depth, lidar), axis=2)
tmp = (tmp * 1000).astype(np.int32) # 米单位*1000保留精度
tmp = SegmentationMapsOnImage(tmp, shape=tmp.shape)
# image, tmp = rsz(image=image, segmentation_maps=tmp)
image, tmp = seq(image=image, segmentation_maps=tmp)
tmp = tmp.arr
tmp = tmp.astype(np.float32) / 1000 # 再转回米
depth, lidar = tmp[:, :, 0], tmp[:, :, 1]
return image, depth, lidar
def myDataAug(X_data, Y_data, its, seed=12):
size = len(X_data)
ia.seed(seed)
seq = iaa.Sequential(
[
iaa.GaussianBlur(
sigma=(0.0, 4.0)), #Random GaussianBlur from sigma 0 to 4
#iaa.AddToHueAndSaturation((-10, 10), per_channel=True), #color jittering
iaa.Affine(translate_percent={
"x": (-0.1, 0.1),
"y": (-0.1, 0.1)
}), #affine transalte (affects segmaps)
iaa.Fliplr(0.5), #50 % to flip horizontally (affects segmaps)
iaa.Flipud(0.5), #50 % to flip vertically (affects segmaps)
iaa.Rotate(
(-45, 45)) # rotate by -45 to 45 degrees (affects segmaps)
],
random_order=True)
X_data_augs = np.zeros((its * size, 1000, 1000, 3), dtype=np.uint8)
Y_data_augs = np.zeros((its * size, 1000, 1000, 3), dtype=np.uint8)
for i in range(its):
X_data_aug, Y_data_aug = seq(images=X_data, segmentation_maps=Y_data)
X_data_augs[i * size:(i + 1) * size] = X_data_aug
Y_data_augs[i * size:(i + 1) * size] = Y_data_aug
return X_data_augs, Y_data_augs
def preprocess(self, ims, augment):
def normalize(batch):
return batch.astype(np.float32) / float(255)
if augment:
augmentations = iaa.Sequential([
iaa.Resize(int(1.1 * hp.img_size)),
iaa.Fliplr(0.5),
iaa.Sometimes(0.4, iaa.Rotate((-30, 30))),
iaa.Sometimes(0.4, iaa.Affine(scale=(0.9, 1.2))),
iaa.Sometimes(0.5,
iaa.PerspectiveTransform(scale=(0.01, 0.20))),
# Crop/resize image to proper dimension
iaa.CropToFixedSize(hp.img_size, hp.img_size),
iaa.Resize(hp.img_size),
iaa.Sometimes(0.3, iaa.SaltAndPepper(0.01)),
iaa.CLAHE(to_colorspace='HSV')
])
else:
augmentations = ia.Sequential([iaa.CLAHE(to_colorspace='HSV')])
augmented = augmentations.augment_images(ims)
for i in augmented:
if i.shape != (hp.img_size, hp.img_size, 3):
print(i.shape)
augmented = np.stack(augmented, axis=0)
return normalize(augmented)
def load_augmentation_aug_geometric():
return iaa.Sequential([
iaa.Sometimes(0.5, iaa.Fliplr()),
iaa.Sometimes(0.5, iaa.Rotate((-45, 45))),
iaa.Sometimes(
0.5,
iaa.Affine(
scale={
"x": (0.5, 1.5),
"y": (0.5, 1.5)
},
order=[0, 1],
mode='constant',
cval=(0, 255),
)),
iaa.Sometimes(
0.5,
iaa.Affine(
translate_percent={
"x": (-0.25, 0.25),
"y": (-0.25, 0.25)
},
order=[0, 1],
mode='constant',
cval=(0, 255),
)),
])
def get_preview(images, augmentationList):
"""
Accepts a list of images and augmentationList as input.
Provides a list of augmented images in that order as ouptut.
"""
augmented = []
for image in images:
for augmentation in augmentationList:
aug_id = augmentation['id']
params = augmentation['params']
if (aug_id == 1):
image = iaa.SaltAndPepper(p=params[0],
per_channel=params[1])(image=image)
elif (aug_id == 2):
image = iaa.imgcorruptlike.GaussianNoise(
severity=(params[0], params[1]))(image=image)
elif (aug_id == 3):
image = iaa.Rain(speed=(params[0], params[1]),
drop_size=(params[2], params[3]))(image=image)
elif (aug_id == 4):
image = iaa.imgcorruptlike.Fog(
severity=(params[0], params[1]))(image=image)
elif (aug_id == 5):
image = iaa.imgcorruptlike.Snow(
severity=(params[0], params[1]))(image=image)
elif (aug_id == 6):
image = iaa.imgcorruptlike.Spatter(
severity=(params[0], params[1]))(image=image)
elif (aug_id == 7):
image = iaa.BlendAlphaSimplexNoise(
iaa.EdgeDetect(1))(image=image)
elif (aug_id == 8):
image = iaa.Rotate(rotate=(params[0], params[1]))(image=image)
elif (aug_id == 9):
image = iaa.Affine()(image=image) #to be implemented
elif (aug_id == 10):
image = iaa.MotionBlur(k=params[0],
angle=(params[1],
params[2]))(image=image)
elif (aug_id == 11):
image = iaa.imgcorruptlike.ZoomBlur(
severity=(params[0], params[1]))(image=image)
elif (aug_id == 12):
image = iaa.AddToBrightness()(image=image) #to be implemented
elif (aug_id == 13):
image = iaa.ChangeColorTemperature(
kelvin=(params[0], params[1]))(image=image)
elif (aug_id == 14):
image = iaa.SigmoidContrast()(image=image) #to be implemented
elif (aug_id == 15):
image = iaa.Cutout(nb_iterations=(params[0], params[1]),
size=params[2],
squared=params[3])(image=image)
else:
print("Not implemented")
augmented.append(image)
return augmented
def get_fg_augmentation():
seq = iaa.Sequential([
iaa.Resize({
'height': 'keep',
'width': (0.9, 1.1)
}),
iaa.KeepSizeByResize(iaa.Rotate((-15, 15), fit_output=True))
])
return seq
def chapter_augmenters_rotate():
fn_start = "geometric/rotate"
image = ia.quokka(size=(128, 128))
aug = iaa.Rotate((-45, 45))
run_and_save_augseq(
fn_start + ".jpg", aug,
[image for _ in range(4*2)], cols=4, rows=2)
def Rotation(self, image, **kwargs):
assert len(image.shape) == 3
image_augs = []
List = [0, 30, 50, 70, 90] # Rotation angle
for angle in List:
seq = iaa.Sequential([iaa.Rotate(rotate=angle)])
image_aug = seq.augment_image(image)
image_augs.append(image_aug)
return image_augs, List
def _compose_transforms(self):
self.aug_functions = list()
for transform_name in self._transforms.keys():
if transform_name == 'resize':
self.aug_functions.append(
iaa.Resize(
{
"height": self._transforms[transform_name].height,
"width": self._transforms[transform_name].width
},
name='resize'))
continue
elif transform_name == 'fliplr':
self.aug_functions.append(
iaa.Fliplr(self._transforms[transform_name].p,
name='fliplr'))
continue
elif transform_name == 'flipud':
self.aug_functions.append(
iaa.Flipud(self._transforms[transform_name].p,
name='flipud'))
continue
elif transform_name == 'rotate':
self.aug_functions.append(
iaa.Rotate(self._transforms[transform_name].degrees,
name='rotate'))
continue
elif transform_name == 'JpegCompression':
self.aug_functions.append(
iaa.JpegCompression(
compression=(self._transforms[transform_name].low,
self._transforms[transform_name].high)))
continue
elif transform_name == 'GaussianBlur':
self.aug_functions.append(
iaa.GaussianBlur(
sigma=(self._transforms[transform_name].low,
self._transforms[transform_name].high)))
continue
elif transform_name == 'CropToFixedSize':
self.aug_functions.append(
iaa.CropToFixedSize(
width=self._transforms[transform_name].width,
height=self._transforms[transform_name].height,
position=self._transforms[transform_name].position))
continue
else:
self.logger.info(
f'{transform_name} is not support in augment build')
self.aug_functions.append(iaa.Noop())
continue
iaa_seq = iaa.Sequential(
self.aug_functions,
name=f'{self.cfg.DATASET.name}_{self.flag}_iaa_seq')
return iaa_seq
def __call__(self, image):
im = image.copy()
pix = np.array(im, dtype='float32')
pix = iaa.Fliplr(0.5).augment_image(pix)
aug_ratate = iaa.OneOf([
iaa.PerspectiveTransform(scale=(0, 0.015)),
iaa.Rotate((-10, 10))
])
pix = aug_ratate.augment_image(pix)
if self.istest:
return pix
pix = self.rand_brightness(pix)
if random.randint(2):
self.aumentone(pix)
#distort = SSD_Compose(self.pd[:-1])
else:
self.aumenttwo(pix)
#distort = SSD_Compose(self.pd[1:])
#pix = distort(pix)
#pix = pix.astype('uint8')
pix = np.where(pix > 255, 255, pix)
pix = np.where(pix < 0, 0, pix)
#im = Image.fromarray(pix, 'RGB')
#self.rand_light_noise(pix)
# aug = iaa.Sequential(iaa.SomeOf(2, [
# iaa.AdditiveGaussianNoise(scale=(0, 0.1*255)),
# iaa.Dropout(p=(0, 0.1), per_channel=0.5),
# iaa.SaltAndPepper(p=(0, 0.1), per_channel=True),
# iaa.GaussianBlur(sigma=(0.0, 3.0)),
# iaa.MotionBlur(k=15, angle=[-45, 45]),
# iaa.AverageBlur(k=((0, 5), (0, 5)))
# ], random_order=True))
aug_noise = iaa.OneOf([
iaa.AdditiveGaussianNoise(scale=(0, 0.1 * 255)),
iaa.Dropout(p=(0, 0.1), per_channel=0.5),
iaa.SaltAndPepper(p=(0, 0.1), per_channel=True),
])
aug_blur = iaa.OneOf([
iaa.GaussianBlur(sigma=(0.0, 1.5)),
iaa.MotionBlur(k=7, angle=[-20, 20]),
iaa.AverageBlur(k=((0, 2.5), (0, 2.5)))
])
pix = aug_noise.augment_image(pix)
pix = aug_blur.augment_image(pix)
return pix / 255.0
def __init__(self):
self.aug = iaa.Sequential([
iaa.Pad(percent=(0, (0, 0.5), 0, (0, 0.5))),
iaa.Resize(size={
"height": 32,
"width": "keep-aspect-ratio"
}),
iaa.Rotate(rotate=(-10, 10)),
])
def make_data_aug(cfg):
transf = iaa.Sequential([
iaa.Flipud(p=0.5),
iaa.Fliplr(p=.5),
iaa.Fliplr(p=.5),
iaa.Rotate(
[360 / cfg.aug_n_angles * i for i in range(cfg.aug_n_angles)])
])
return transf
def __init__(self):
self.augmentations = augmenters.Sequential([
augmenters.Sometimes(0.5, augmenters.Flipud(0.5)),
augmenters.Sometimes(0.5, augmenters.Fliplr(0.5)),
augmenters.Sometimes(0.5, augmenters.Rotate((-45, 45))),
augmenters.Sometimes(0.25, augmenters.ShearX((-20, 20))),
augmenters.Sometimes(
0.1, augmenters.AdditiveGaussianNoise(scale=(0, 0.1 * 255)))
],
random_order=True)
def augmentation():
train_images_horiz, train_images_verti, train_labels, test_images_horiz, test_images_verti, test_labels, val_images_horiz, val_images_verti, val_labels = list_input_network(
)
seq1 = iaa.Sequential([iaa.Fliplr(1)])
seq2 = iaa.Sequential([iaa.Flipud(1)])
seq3 = iaa.Sequential([iaa.Rotate((-45, 45))])
seq4 = iaa.Sequential([iaa.TranslateX(px=(-20, 20))])
seq5 = iaa.Sequential([iaa.CropToFixedSize(width=30, height=30)])
new_train_images = train_images_horiz
new_train_labels = train_labels
""" different sequences of data augmentation applied """
images_aug = seq1(images=train_images_horiz)
new_train_images = np.concatenate((new_train_images, images_aug))
new_train_labels = np.concatenate((new_train_labels, train_labels))
images_aug = seq2(images=train_images_horiz)
new_train_images_horiz = np.concatenate((new_train_images, images_aug))
new_train_labels = np.concatenate((new_train_labels, train_labels))
images_aug = seq3(images=train_images_horiz)
new_train_images = np.concatenate((new_train_images, images_aug))
new_train_labels = np.concatenate((new_train_labels, train_labels))
images_aug = seq4(images=train_images_horiz)
new_train_images_horiz = np.concatenate((new_train_images, images_aug))
new_train_labels = np.concatenate((new_train_labels, train_labels))
#new_train_images = train_images_verti
""" same on vertical images """
"""
images_aug = seq1(images=train_images_verti)
new_train_images = np.concatenate((new_train_images, images_aug))
images_aug = seq2(images=train_images_verti)
new_train_images_verti = np.concatenate((new_train_images, images_aug))
images_aug = seq3(images=train_images_verti)
new_train_images = np.concatenate((new_train_images, images_aug))
images_aug = seq4(images=train_images_verti)
new_train_images_verti = np.concatenate((new_train_images, images_aug))
"""
# images_aug = seq5(images=train_images)
# new_train_images = np.concatenate((new_train_images, images_aug))
# new_train_labels = np.concatenate((new_train_labels, train_labels))
#return new_train_images_horiz, new_train_images_verti, new_train_labels
return new_train_images_horiz, new_train_labels
def __call__(self, *args, **kwargs) -> typing.Tuple[np.ndarray, typing.List[Polygon]]:
if self.is_training:
resize = iaa.Resize(size=dict(longer_side=self.long_sizes,
width='keep-aspect-ratio'))
rotate = iaa.Rotate(rotate=self.angles, fit_output=True)
resize_height = iaa.Resize(size=dict(height=self.height_ratios,
width='keep'))
crop = iaa.CropToFixedSize(width=self.cropped_size[0], height=self.cropped_size[1])
fix_resize = iaa.Resize(size=self.output_size)
# blur = iaa.GaussianBlur()
# blur = iaa.Sometimes(p=self.blur_prob,
# then_list=blur)
brightness = iaa.MultiplyBrightness((0.5, 1.5))
brightness = iaa.Sometimes(self.color_jitter_prob, then_list=brightness)
saturation = iaa.MultiplySaturation((0.5, 1.5))
saturation = iaa.Sometimes(self.color_jitter_prob, then_list=saturation)
contrast = iaa.LinearContrast(0.5)
contrast = iaa.Sometimes(self.color_jitter_prob, then_list=contrast)
hue = iaa.MultiplyHue()
hue = iaa.Sometimes(self.color_jitter_prob, then_list=hue)
augs = [resize,
rotate,
resize_height,
crop,
fix_resize,
brightness,
saturation,
contrast,
hue]
ia = iaa.Sequential(augs)
else:
fix_resize = iaa.Resize(size=self.output_size)
ia = iaa.Sequential([fix_resize])
image = args[0]
polygons = args[1]
polygon_list = []
for i in range(polygons.shape[0]):
polygon_list.append(Polygon(polygons[i].tolist()))
polygons_on_image = PolygonsOnImage(polygon_list, shape=image.shape)
image_aug, polygons_aug = ia(image=image, polygons=polygons_on_image)
return image_aug, polygons_aug.polygons
def dictRotateAug(baseImageListFunc, baseMaskListFunc, fullImageListFunc, segmapListFunc):
print('Rotation, starting number of images:', len(segmapListFunc))
rotateAug_x00percent = 2
rotateAug = iaa.Rotate((-3, 3), mode="reflect")
rotateAug._mode_segmentation_maps = "reflect"
alteredImageListFunc, alteredMaskListFunc = expandList(baseImageListFunc, baseMaskListFunc, rotateAug_x00percent)
(alteredImageListFunc, alteredMaskListFunc) = rotateAug(images=alteredImageListFunc,
segmentation_maps=alteredMaskListFunc)
fullImageListFunc.extend(alteredImageListFunc)
segmapListFunc.extend(alteredMaskListFunc)
return fullImageListFunc, segmapListFunc
def __init__(self, iaalist=None):
if iaalist is None:
iaalist = iaa.Sequential([
iaa.Sometimes(0.5, iaa.ChannelShuffle(0.3)),
iaa.Sometimes(0.5, iaa.MultiplyHue((0.5, 1.5))),
iaa.Sometimes(0.5, iaa.AddToHueAndSaturation((-50, 50), per_channel=True)),
iaa.Sometimes(0.5, iaa.Fliplr(0.5)),
iaa.Sometimes(0.5, iaa.Flipud(0.5)),
iaa.Sometimes(0.5, iaa.Rotate((-50, 50)))
], random_order=True)
self.transformSet = iaalist
self.outscale = random.choice([0.8, 0.85, 0.9, 0.95])
def rotate(magnitude: int) -> iaa.BlendAlphaBoundingBoxes:
"""
Rotate the bounding box in an image
Tensorflow Policy Equivalent: rotate_with_bboxes
:type magnitude: int
:param magnitude: magnitude of rotation
:rtype: iaa.BlendAlphaBoundingBoxes
:return: Method to apply rotation
"""
level = _rotate_mag_to_arg(magnitude)
return iaa.Rotate(level)
def make_rotation_ops(increments):
ops = []
for increment in increments:
suffix = f"{increment}"
if increment < 0:
suffix = f"Back{-increment}"
ops.append(
Augment("Rotate" + suffix, iaa.Sequential([
iaa.Grayscale(alpha=1.0),
iaa.Rotate(increment)
]), num_repetitions=1)
)
return ops
def gen_steps():
STEPS=[
iaa.Rotate((-50,-20)),
iaa.Rotate((20,50)),
iaa.Affine(shear=(-30,-10)),
iaa.Affine(shear=(10,30)),
iaa.ShearX(((-10, 10))),
iaa.ShearY(((-10, 10))),
iaa.Affine(scale=(0.5, 1.5)),
iaa.Affine(scale=(1, 2)),
iaa.Affine(translate_px={"x": (1, random.randint(300,800)), "y": (1, random.randint(300,800))}),
[
iaa.Affine(translate_px={"x": (1, random.randint(300,800)), "y": (1, random.randint(300,800))}),
iaa.Affine(shear=(-15,15)),
],
[
iaa.Affine(translate_px={"x": (1, random.randint(300,800)), "y": (1, random.randint(300,800))}),
iaa.Affine(shear=(-15,15)),
],
]
return STEPS
def __init__(self):
sometimes = lambda aug: iaa.Sometimes(0.5, aug)
self.aug = iaa.SomeOf(
3,
[
sometimes(
iaa.AddToHueAndSaturation((-50, 50), per_channel=True)),
sometimes(
iaa.SigmoidContrast(
gain=(3, 10), cutoff=(0.4, 0.6), per_channel=True)),
#sometimes(iaa.PiecewiseAffine(scale=(0.01, 0.02))),
sometimes(iaa.Invert(0.7, per_channel=0.8)),
sometimes(iaa.Rotate((-20, 20)))
])
def train(model):
"""Train the model."""
# Training dataset.
dataset_train = CharacterDataset()
dataset_train.load_characters("train")
dataset_train.prepare()
# Validation dataset
dataset_val = CharacterDataset()
dataset_val.load_characters("val")
dataset_val.prepare()
#Augmentation
aug = iaa.SomeOf(2, [
iaa.AdditiveGaussianNoise(scale=(0, 0.10 * 255)),
iaa.MotionBlur(),
iaa.GaussianBlur(sigma=(0.0, 2.0)),
iaa.RemoveSaturation(mul=(0, 0.5)),
iaa.GammaContrast(),
iaa.Rotate(rotate=(-45, 45)),
iaa.PerspectiveTransform(scale=(0.01, 0.15)),
iaa.JpegCompression(compression=(0, 75)),
iaa.imgcorruptlike.Spatter(severity=(1, 4)),
iaa.Rain(speed=(0.1, 0.3)),
iaa.Fog()
])
custom_callbacks = [
ReduceLROnPlateau(monitor='val_loss',
factor=0.1,
patience=5,
verbose=1),
EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1)
]
# *** This training schedule is an example. Update to your needs ***
# Since we're using a very small dataset, and starting from
# COCO trained weights, we don't need to train too long. Also,
# no need to train all layers, just the heads should do it.
print("Training network heads")
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=100,
layers='heads',
augmentation=aug,
custom_callbacks=custom_callbacks)
def balanced_augmentation(X, Y, epochs, img_size):
print("START AUGMENTATION...")
cluster_df = pd.read_csv('cluster_df.csv', usecols=['cluster'])
cluster_size = [0, 0, 0]
cluster_size[0] = cluster_df.value_counts()[0].iloc[0]
cluster_size[1] = cluster_df.value_counts()[1].iloc[0]
cluster_size[2] = cluster_df.value_counts()[2].iloc[0]
d_len = cluster_size[0] * epochs[0] + cluster_size[1] * epochs[
1] + cluster_size[2] * epochs[2]
augmented_x = np.zeros((d_len, img_size, img_size, 3), dtype=np.float)
augmented_y = np.zeros((d_len, img_size, img_size, 1), dtype=np.bool)
seq = iaa.Sequential([
iaa.Fliplr(0.5),
iaa.GaussianBlur(sigma=(0, 3.0)),
iaa.Rotate((-40, 40))
])
global_cnt = 0
for i in range(X.shape[0]):
img_cluster = cluster_df.iloc[i]['cluster']
'''if img_cluster == 2:
print(epochs[img_cluster])
cv2.imshow("img", X[i])
cv2.waitKey(0)
cv2.destroyAllWindows()'''
for e in range(epochs[img_cluster]):
img, seg = seq(images=[X[i]], segmentation_maps=[Y[i]])
augmented_x[global_cnt] = img[0]
augmented_y[global_cnt] = seg[0]
global_cnt += 1
print("AUGMENTED")
return augmented_x, augmented_y
def do_all_aug(image):
do_aug(image, iaa.Noop(name="origin"))
do_aug(image, iaa.Crop((0, 10))) # 切边
do_aug(image, iaa.GaussianBlur((0, 3)))
do_aug(image, iaa.AverageBlur(1, 7))
do_aug(image, iaa.MedianBlur(1, 7))
do_aug(image, iaa.Sharpen())
do_aug(image, iaa.BilateralBlur()) # 既噪音又模糊，叫双边
do_aug(image, iaa.MotionBlur())
do_aug(image, iaa.MeanShiftBlur())
do_aug(image, iaa.GammaContrast())
do_aug(image, iaa.SigmoidContrast())
do_aug(image,
iaa.Affine(shear={
'x': (-10, 10),
'y': (-10, 10)
}, mode="edge")) # shear：x轴往左右偏离的像素书，(a,b)是a,b间随机值，[a,b]是二选一
do_aug(image,
iaa.Affine(shear={
'x': (-10, 10),
'y': (-10, 10)
}, mode="edge")) # shear：x轴往左右偏离的像素书，(a,b)是a,b间随机值，[a,b]是二选一
do_aug(image, iaa.Rotate(rotate=(-10, 10), mode="edge"))
do_aug(image, iaa.PiecewiseAffine()) # 局部点变形
do_aug(image, iaa.Fog())
do_aug(image, iaa.Clouds())
do_aug(image, iaa.Snowflakes(flake_size=(0.1, 0.2),
density=(0.005, 0.025)))
do_aug(
image,
iaa.Rain(
nb_iterations=1,
drop_size=(0.05, 0.1),
speed=(0.04, 0.08),
))
do_aug(
image,
iaa.ElasticTransformation(alpha=(0.0, 20.0),
sigma=(3.0, 5.0),
mode="nearest"))
do_aug(image, iaa.AdditiveGaussianNoise(scale=(0, 10)))
do_aug(image, iaa.AdditiveLaplaceNoise(scale=(0, 10)))
do_aug(image, iaa.AdditivePoissonNoise(lam=(0, 10)))
do_aug(image, iaa.Salt((0, 0.02)))
do_aug(image, iaa.Pepper((0, 0.02)))
def train(model):
"""Train the model."""
# Training dataset.
dataset_train = PlateDataset()
dataset_train.load_plates("train")
dataset_train.prepare()
# Validation dataset
dataset_val = PlateDataset()
dataset_val.load_plates("val")
dataset_val.prepare()
#Augmentation
aug = iaa.OneOf([
iaa.GaussianBlur(sigma=(0, 1.0)),
iaa.MotionBlur(),
iaa.RemoveSaturation((0.0, 0.5)),
iaa.GammaContrast(),
iaa.Rotate(rotate=(-45, 45)),
iaa.PerspectiveTransform(scale=(0.01, 0.15)),
iaa.SaltAndPepper(),
iaa.JpegCompression(compression=(0, 75)),
iaa.imgcorruptlike.Spatter(severity=(1, 4)),
iaa.imgcorruptlike.DefocusBlur(severity=1)
])
custom_callbacks = [
ReduceLROnPlateau(monitor='val_loss',
factor=0.1,
patience=5,
verbose=1),
EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1)
]
# Since we're using a very small dataset, and starting from
# COCO trained weights, we don't need to train too long. Also,
# no need to train all layers, just the heads should do it.
print("Training network heads")
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=100,
layers='all',
augmentation=aug,
custom_callbacks=custom_callbacks)
