Image Perturbators¶
All image-specific perturbators derive from the BlackboxImagePerturbator class. They take an PIL image as input and
return a transformed version of it.
ImageNoisePerturbator¶
The ImageNoisePerturbator adds noise to an image. The probability distribution determines how much noise is added.
Usage¶
from insynth.perturbators.image import ImageNoisePerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageNoisePerturbator(noise_prob=norm, noise_prob_args={'loc': 0.01, 'scale': 0.005})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageBrightnessPerturbator¶
The ImageBrightnessPerturbator changes the brightness of an image. The probability distribution determines how much the brightness is changed.
Usage¶
from insynth.perturbators.image import ImageBrightnessPerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageBrightnessPerturbator(brightness_change_prob=norm,
brightness_change_prob_args={'loc': 1, 'scale': 0.5})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageContrastPerturbator¶
The ImageContrastPerturbator changes the contrast of an image. The probability distribution determines how much the contrast is changed.
Usage¶
from insynth.perturbators.image import ImageContrastPerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageContrastPerturbator(contrast_change_prob=norm, contrast_change_prob_args={'loc': 1, 'scale': 0.5})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageSharpnessPerturbator¶
The ImageSharpnessPerturbator changes the sharpness of an image. The probability distribution determines how much the sharpness is changed.
Usage¶
from insynth.perturbators.image import ImageSharpnessPerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageSharpnessPerturbator(sharpness_change_prob=norm, sharpness_change_prob_args={'loc': 1, 'scale': 0.5})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageFlipPerturbator¶
The ImageFlipPerturbator flips an image horizontally or vertically. There is no probability distribution for this perturbator.
Usage¶
from insynth.perturbators.image import ImageFlipPerturbator
from PIL import Image
perturbator = ImageFlipPerturbator()
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageOcclusionPerturbator¶
The ImageOcclusionPerturbator adds occlusion to an image. The probability distribution determines how much occlusion is added. Also, a second and third probability distribution determines the size of each occlusion.
Usage¶
from insynth.perturbators.image import ImageOcclusionPerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageOcclusionPerturbator(strength_prob=norm, strength_prob_args={'loc': 0.01, 'scale': 0.005},
width_prob=norm,
width_prob_args={'loc': 10, 'scale': 5}, height_prob=norm,
height_prob_args={'loc': 10, 'scale': 5}, color='#000000')
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImageCompressionPerturbator¶
The ImageCompressionPerturbator adds compression artefacts to an image. The probability distribution determines how much compression artefacts are added.
Usage¶
from insynth.perturbators.image import ImageCompressionPerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImageCompressionPerturbator(artifact_prob=norm, artifact_prob_args={'loc': 0.5, 'scale': 0.2})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
ImagePixelizePerturbator¶
The ImagePixelizePerturbator pixelates an image. The probability distribution determines how much pixelation is added.
Usage¶
from insynth.perturbators.image import ImagePixelizePerturbator
from PIL import Image
from scipy.stats import norm
perturbator = ImagePixelizePerturbator(pixelize_prob=norm, pixelize_prob_args={'loc': 0.2, 'scale': 0.1})
im = Image.open('image.jpg')
perturbed_im = perturbator.apply(im)
Result¶
