mirror of
https://github.com/AntoineHX/smart_augmentation.git
synced 2025-05-04 12:10:45 +02:00
Minor improvement + Comments
This commit is contained in:
Harle, Antoine (Contracteur)
parent
d21a6bbf5c
commit
c1ad787d97
5 changed files with 165 additions and 62 deletions
|
|
@ -53,64 +53,91 @@ TF_dict={ #Dataugv5 #AutoAugment
|
|||
#'Equalize': (lambda mag: None),
|
||||
}
|
||||
'''
|
||||
# Dictionnary mapping tranformations identifiers to their function.
|
||||
# Each value of the dict should be a lambda function taking a (batch of data, magnitude of transformations) tuple as input and returns a batch of data.
|
||||
TF_dict={ #Dataugv5
|
||||
## Geometric TF ##
|
||||
'Identity' : (lambda x, mag: x),
|
||||
'FlipUD' : (lambda x, mag: flipUD(x)),
|
||||
'FlipLR' : (lambda x, mag: flipLR(x)),
|
||||
'Rotate': (lambda x, mag: rotate(x, angle=rand_floats(size=x.shape[0], mag=mag, maxval=30))),
|
||||
'TranslateX': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=20), zero_pos=0))),
|
||||
'TranslateY': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=20), zero_pos=1))),
|
||||
'ShearX': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=0.3), zero_pos=0))),
|
||||
'ShearY': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=0.3), zero_pos=1))),
|
||||
## Geometric TF ##
|
||||
'Identity' : (lambda x, mag: x),
|
||||
'FlipUD' : (lambda x, mag: flipUD(x)),
|
||||
'FlipLR' : (lambda x, mag: flipLR(x)),
|
||||
'Rotate': (lambda x, mag: rotate(x, angle=rand_floats(size=x.shape[0], mag=mag, maxval=30))),
|
||||
'TranslateX': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=20), zero_pos=0))),
|
||||
'TranslateY': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=20), zero_pos=1))),
|
||||
'ShearX': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=0.3), zero_pos=0))),
|
||||
'ShearY': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, maxval=0.3), zero_pos=1))),
|
||||
|
||||
## Color TF (Expect image in the range of [0, 1]) ##
|
||||
'Contrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Color':(lambda x, mag: color(x, color_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Brightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Posterize': (lambda x, mag: posterize(x, bits=rand_floats(size=x.shape[0], mag=mag, minval=4., maxval=8.))),#Perte du gradient
|
||||
'Solarize': (lambda x, mag: solarize(x, thresholds=rand_floats(size=x.shape[0], mag=mag, minval=1/256., maxval=256/256.))), #Perte du gradient #=>Image entre [0,1]
|
||||
## Color TF (Expect image in the range of [0, 1]) ##
|
||||
'Contrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Color':(lambda x, mag: color(x, color_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Brightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.9))),
|
||||
'Posterize': (lambda x, mag: posterize(x, bits=rand_floats(size=x.shape[0], mag=mag, minval=4., maxval=8.))),#Perte du gradient
|
||||
'Solarize': (lambda x, mag: solarize(x, thresholds=rand_floats(size=x.shape[0], mag=mag, minval=1/256., maxval=256/256.))), #Perte du gradient #=>Image entre [0,1]
|
||||
|
||||
#Color TF (Common mag scale)
|
||||
'+Contrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Color':(lambda x, mag: color(x, color_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Brightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'-Contrast': (lambda x, mag: contrast(x, contrast_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Color':(lambda x, mag: color(x, color_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Brightness':(lambda x, mag: brightness(x, brightness_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'=Posterize': (lambda x, mag: posterize(x, bits=invScale_rand_floats(size=x.shape[0], mag=mag, minval=4., maxval=8.))),#Perte du gradient
|
||||
'=Solarize': (lambda x, mag: solarize(x, thresholds=invScale_rand_floats(size=x.shape[0], mag=mag, minval=1/256., maxval=256/256.))), #Perte du gradient #=>Image entre [0,1]
|
||||
|
||||
'BShearX': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=0.3*3, maxval=0.3*4), zero_pos=0))),
|
||||
'BShearY': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=0.3*3, maxval=0.3*4), zero_pos=1))),
|
||||
'BTranslateX': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=25, maxval=30), zero_pos=0))),
|
||||
'BTranslateX-': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=-25, maxval=-30), zero_pos=0))),
|
||||
'BTranslateY': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=25, maxval=30), zero_pos=1))),
|
||||
'BTranslateY-': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=-25, maxval=-30), zero_pos=1))),
|
||||
|
||||
'BadContrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.9*2, maxval=2*4))),
|
||||
'BadBrightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.9, maxval=2*3))),
|
||||
#Color TF (Common mag scale)
|
||||
'+Contrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Color':(lambda x, mag: color(x, color_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Brightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'+Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.0, maxval=1.9))),
|
||||
'-Contrast': (lambda x, mag: contrast(x, contrast_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Color':(lambda x, mag: color(x, color_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Brightness':(lambda x, mag: brightness(x, brightness_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'-Sharpness':(lambda x, mag: sharpeness(x, sharpness_factor=invScale_rand_floats(size=x.shape[0], mag=mag, minval=0.1, maxval=1.0))),
|
||||
'=Posterize': (lambda x, mag: posterize(x, bits=invScale_rand_floats(size=x.shape[0], mag=mag, minval=4., maxval=8.))),#Perte du gradient
|
||||
'=Solarize': (lambda x, mag: solarize(x, thresholds=invScale_rand_floats(size=x.shape[0], mag=mag, minval=1/256., maxval=256/256.))), #Perte du gradient #=>Image entre [0,1]
|
||||
|
||||
'Random':(lambda x, mag: torch.rand_like(x)),
|
||||
'RandBlend': (lambda x, mag: blend(x,torch.rand_like(x), alpha=torch.tensor(0.7,device=mag.device).expand(x.shape[0]))),
|
||||
|
||||
#Non fonctionnel
|
||||
#'Auto_Contrast': (lambda mag: None), #Pas opti pour des batch (Super lent)
|
||||
#'Equalize': (lambda mag: None),
|
||||
## Bad Tranformations ##
|
||||
# Bad Geometric TF #
|
||||
'BShearX': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=0.3*3, maxval=0.3*4), zero_pos=0))),
|
||||
'BShearY': (lambda x, mag: shear(x, shear=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=0.3*3, maxval=0.3*4), zero_pos=1))),
|
||||
'BTranslateX': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=25, maxval=30), zero_pos=0))),
|
||||
'BTranslateX-': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=-25, maxval=-30), zero_pos=0))),
|
||||
'BTranslateY': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=25, maxval=30), zero_pos=1))),
|
||||
'BTranslateY-': (lambda x, mag: translate(x, translation=zero_stack(rand_floats(size=(x.shape[0],), mag=mag, minval=-25, maxval=-30), zero_pos=1))),
|
||||
|
||||
# Bad Color TF #
|
||||
'BadContrast': (lambda x, mag: contrast(x, contrast_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.9*2, maxval=2*4))),
|
||||
'BadBrightness':(lambda x, mag: brightness(x, brightness_factor=rand_floats(size=x.shape[0], mag=mag, minval=1.9, maxval=2*3))),
|
||||
|
||||
# Random TF #
|
||||
'Random':(lambda x, mag: torch.rand_like(x)),
|
||||
'RandBlend': (lambda x, mag: blend(x,torch.rand_like(x), alpha=torch.tensor(0.7,device=mag.device).expand(x.shape[0]))),
|
||||
|
||||
#Non fonctionnel
|
||||
#'Auto_Contrast': (lambda mag: None), #Pas opti pour des batch (Super lent)
|
||||
#'Equalize': (lambda mag: None),
|
||||
}
|
||||
|
||||
TF_no_mag={'Identity', 'FlipUD', 'FlipLR', 'Random', 'RandBlend'}
|
||||
TF_no_grad={'Solarize', 'Posterize', '=Solarize', '=Posterize'}
|
||||
TF_ignore_mag= TF_no_mag | TF_no_grad
|
||||
TF_no_mag={'Identity', 'FlipUD', 'FlipLR', 'Random', 'RandBlend'} #TF that don't have use for magnitude parameter.
|
||||
TF_no_grad={'Solarize', 'Posterize', '=Solarize', '=Posterize'} #TF which implemetation doesn't allow gradient propagaition.
|
||||
TF_ignore_mag= TF_no_mag | TF_no_grad #TF for which magnitude should be ignored (Magnitude fixed).
|
||||
|
||||
def int_image(float_image): #ATTENTION : legere perte d'info (granularite : 1/256 = 0.0039)
|
||||
return (float_image*255.).type(torch.uint8)
|
||||
def int_image(float_image):
|
||||
"""Convert a float Tensor/Image to an int Tensor/Image.
|
||||
|
||||
Be warry that this transformation isn't bijective, each conversion will result in small loss of information.
|
||||
Granularity: 1/256 = 0.0039.
|
||||
|
||||
This will also result in the loss of the gradient associated to input as gradient cannot be tracked on int Tensor.
|
||||
|
||||
Args:
|
||||
float_image (torch.float): Image tensor.
|
||||
|
||||
Returns:
|
||||
(torch.uint8) Converted tensor.
|
||||
"""
|
||||
return (float_image*255.).type(torch.uint8)
|
||||
|
||||
def float_image(int_image):
|
||||
return int_image.type(torch.float)/255.
|
||||
"""Convert a int Tensor/Image to an float Tensor/Image.
|
||||
|
||||
Args:
|
||||
int_image (torch.uint8): Image tensor.
|
||||
|
||||
Returns:
|
||||
(torch.float) Converted tensor.
|
||||
"""
|
||||
return int_image.type(torch.float)/255.
|
||||
|
||||
#def rand_inverse(value):
|
||||
# return value if random.random() < 0.5 else -value
|
||||
|
|
@ -125,11 +152,22 @@ def float_image(int_image):
|
|||
# if not minval : minval = -real_max
|
||||
# return random.uniform(minval, real_max)
|
||||
|
||||
def rand_floats(size, mag, maxval, minval=None): #[(-maxval,minval), maxval]
|
||||
real_mag = float_parameter(mag, maxval=maxval)
|
||||
if not minval : minval = -real_mag
|
||||
#return random.uniform(minval, real_max)
|
||||
return minval + (real_mag-minval) * torch.rand(size, device=mag.device) #[min_val, real_mag]
|
||||
def rand_floats(size, mag, maxval, minval=None):
|
||||
"""Generate a batch of random values.
|
||||
|
||||
Args:
|
||||
size (int): Number of value to generate.
|
||||
mag (float): Level of the operation that will be between [PARAMETER_MIN, PARAMETER_MAX].
|
||||
maxval (float): Maximum value that can be generated. This will be scaled to mag/PARAMETER_MAX.
|
||||
minval (float): Minimum value that can be generated. (default: -maxval)
|
||||
|
||||
Returns:
|
||||
Generated batch of float values between [minval, maxval].
|
||||
"""
|
||||
real_mag = float_parameter(mag, maxval=maxval)
|
||||
if not minval : minval = -real_mag
|
||||
#return random.uniform(minval, real_max)
|
||||
return minval + (real_mag-minval) * torch.rand(size, device=mag.device) #[min_val, real_mag]
|
||||
|
||||
def invScale_rand_floats(size, mag, maxval, minval):
|
||||
#Mag=[0,PARAMETER_MAX] => [PARAMETER_MAX, 0] = [maxval, minval]
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue