Augmentor

last updated on 10/10/2023

Paper

Supported languages:

Python (2.7, 3.5, 3.6, 3.7, 3.8, 3.9)
Julia

Supported ML libraries:

PyTorch
Keras
Flux

Augmentor aims to be a standalone library that is platform and framework independent. It employs a stochastic approach using building blocks that allow for operations to be pieced together in a pipeline.

Augmentor is written in Python. A Julia version of the package is also being developed as a sister project.

Transformations Description Tags

It crops the images based on the passed parameters (width and height of the area to be cropped and whether the area to be cropped should be taken from the center of the image or from a random position within the image).

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Crop(probability=1, width=1000, height=1000, centre=True)`

Original	Augmented

`operation = A.Operations.Crop(probability=1, width=1000, height=1000, centre=False)`

Crops images by a percentage of their area.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.CropPercentage(probability=1, percentage_area=0.7, centre=False, randomise_percentage_area=False)`

It crops a random area from an image, based on the area specified in the percent_area parameter.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.CropRandom(probability=1, percentage_area=0.8)`

Performs randomised, elastic distortions on images.

DISTORTION

Original	Augmented

`operation = A.Operations.Distort(probability=1, grid_width=21, grid_height=21, magnitude=10)`

Original	Augmented

`operation = A.Operations.Distort(probability=1, grid_width=21, grid_height=21, magnitude=20)`

Allows an image to be mirrored along either its x axis or its y axis, or randomly.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Flip(probability=1, top_bottom_left_right='LEFT_RIGHT')`

Original	Augmented

`operation = A.Operations.Flip(probability=1, top_bottom_left_right='TOP_BOTTOM')`

Performs randomised, elastic gaussian distortions on images.

DISTORTION

Original	Augmented

`operation = A.Operations.GaussianDistortion(probability=1, grid_width=50, grid_height=50, magnitude=5, corner='ur', method='in', mex=0.9, mey=0.9, sdx=1, sdy=1)`

Original	Augmented

`operation = A.Operations.GaussianDistortion(probability=1, grid_width=50, grid_height=50, magnitude=5, corner='ur', method='out', mex=0.9, mey=0.9, sdx=1, sdy=1)`

Original	Augmented

`operation = A.Operations.GaussianDistortion(probability=1, grid_width=50, grid_height=50, magnitude=5, corner='ur', method='out', mex=0.1, mey=0.9, sdx=1, sdy=1)`

Original	Augmented

`operation = A.Operations.GaussianDistortion(probability=1, grid_width=50, grid_height=50, magnitude=5, corner='ur', method='in', mex=0.5, mey=0.5, sdx=1.5, sdy=1.5)`

Converts images into greyscale.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.Greyscale(probability=1)`

Performs histogram equalisation.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.HistogramEqualisation(probability=1)`

Inverts the values of the images.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.Invert(probability=1)`

10.

It randomly modifies (within given parameters) the brightness of the image.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.RandomBrightness(probability=1, min_factor=0.5, max_factor=0.9)`

Original	Augmented

`operation = A.Operations.RandomBrightness(probability=1, min_factor=0.9, max_factor=1.5)`

11.

It randomly modifies (within given parameters) the saturation of the image.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.RandomColor(probability=1, min_factor=0.4, max_factor=0.7)`

Original	Augmented

`operation = A.Operations.RandomColor(probability=1, min_factor=1.2, max_factor=1.6)`

12.

It randomly modifies (within given parameters) the contrast of the image.

PHOTOMETRY

Original	Augmented

`operation = A.Operations.RandomContrast(probability=1, min_factor=0.3, max_factor=0.6)`

Original	Augmented

`operation = A.Operations.RandomContrast(probability=1, min_factor=1.4, max_factor=1.5)`

13.

Performs a Random Erasing.

DROPOUT

Original	Augmented

`operation = A.Operations.RandomErasing(probability=1, rectangle_area=0.4)`

Original	Augmented

`operation = A.Operations.RandomErasing(probability=1, rectangle_area=1)`

14.

Resizes images by absolute values passed as parameters.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Resize(probability=1, width=380, height=300, resample_filter='NEAREST')`

15.

Performs rotations on images in multiples of 90 degrees. The rotation parameter must be one of 90, 180, 270 or -1 (rotates the image randomly by either 90, 180, or 270 degrees).

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Rotate(probability=1, rotation=180)`

16.

Performs rotations. The returned image has the same dimensions as the original one: after the rotation has been performed, the largest possible area of the same aspect ratio of the original image is cropped from the rotated image, and this is then resized to match the original image size.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.RotateRange(probability=1, max_left_rotation=1, max_right_rotation=45)`

17.

Perform rotations without automatically cropping the images, as opposed to the RotateRange class.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.RotateStandard(probability=1, max_left_rotation=1, max_right_rotation=90, expand=False, fillcolor=None)`

18.

This class is used to increase or decrease images in size by a certain factor, while maintaining the aspect ratio of the original image.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Scale(probability=1, scale_factor=0.1)`

19.

Shears the images in a certain direction. The shearing can occur along the x or y axis and in both directions (i.e., left or right along the x axis, up or down along the y axis). This transformation returns an image of the same dimension as the input image. The shearing axis is randomized.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Shear(probability=1, max_shear_left=25, max_shear_right=1)`

20.

Performs perspective skewing.

BASIC GEOMETRY

Original	Augmented

`# with numpy==1.19.5 operation = A.Operations.Skew(probability=1, skew_type='TILT', magnitude=5)`

Original	Augmented

`# with numpy==1.19.5 operation = A.Operations.Skew(probability=1, skew_type='CORNER', magnitude=5)`

21.

Zooms the images but returns a cropped region of the zoomed image of the same dimension as the original image. The amount of applied zoom is random, between min_factor and max_factor.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.Zoom(probability=1, min_factor=2, max_factor=2)`

22.

Zooms the images but returns a cropped region of the zoomed image of the same dimension as the original image. The amount of applied zoom is random, between min_factor and max_factor.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.ZoomGroundTruth(probability=1, min_factor=2, max_factor=2)`

23.

Zooms into random areas of the image. Zooms into a random area of the image, rather than the centre of the image, as is done by Zoom. The zoom factor is fixed unless randomise is set to True.

BASIC GEOMETRY

Original	Augmented

`operation = A.Operations.ZoomRandom(probability=1, percentage_area=0.8, randomise=False)`

	Description	Tags
1.	Adjusts the brightness and contrast of an image according to the formula `α * image[i] + β * M`, where `M` is either `mean(image)` or the maximum intensity value.
2.	Combines the first dimension of a given array into a colorant of type `colortype` using the function `ImageCore.colorview`. The main difference is that a separate color channel is also expected for Gray images.
3.	Converts the element type of the given array/image into the given eltype. This operation is especially useful for converting color images to grayscale (or the other way around). That said, the operation is not specific to color types and can also be used for numeric arrays (e.g. with separated channels).
4.	Crops out the area denoted by the specified pixel ranges.	CROP
5.	Crops out the area denoted by the specified pixel ranges. In contrast to Crop, the position `x=1` `y=1` is not necessarily located at the top left of the current image, but instead depends on the cumulative effect of the previous transformations.	CROP
6.	Crops out the biggest area around the center of the given image such that the output image satisfies the specified aspect ratio.	CROP
7.	Crops out the area of the specified pixel size around the center of the input image.	CROP
8.	Distorts the given image using a randomly (uniform) generated vector field of the given grid size.
9.	Horizontally flips the input image.	FLIP
10.	Vertically flips the input image.	FLIP
11.	Blurs an image using a Gaussian filter.	BLUR
12.	Crops out the biggest possible area at some random position of the given image, such that the output image satisfies the specified aspect ratio.	CROP
13.	Crops out an area of predefined size at some random position of the given image.	CROP
14.	Resizes the image to a pre-specified fixed size; the aspect ratio of the image is not preserved.
15.	Rotates counterclockwise around the center of the image.	ROTATION
16.	Rotates the image by 180 degrees.	ROTATION
17.	Rotates the image by 270 degrees.	ROTATION
18.	Rotates the image by 90 degrees.	ROTATION
19.	Resizes the image by multiplying the height and width by specified factors; the output image size depends on the size of the input image.
20.	Performs horizontal shear.	SHEAR
21.	Performs vertical shear.	SHEAR
22.	Splits out the color channels of the given image.
23.	Resizes the height and width of the image based on specified factors, but crops the image in a way that maintains the original dimensions.

Libraries

Augmentor