CVAugmentor

The CVAugmentor Augmentation Package.

CVAugmentor is a python package for augmenting image and video files. It provides a number of utilities that aid augmentation in a automated manner. The aim of the package is to make augmentation easier. The package is built on top of the OpenCV and Pillow libraries.

class CVAugmentor.pipeline.Pipeline

augment(input_path: str, output_path: str, target: str, process_type: str, mode: str, augmentations: Dict[str, Callable], verbose: Optional[bool] = False, aug_verbose: Optional[bool] = False, warn_verbose: Optional[bool] = False) → None

Augments the input data.

Parameters

• input_path (str) – Path to the input data.

• output_path (str) – Path to the output data.

• target (str) –

  Target of the augmentation.

  The options are:

  ”image”

  Specifies that the target of the augmentation is an image.

  ”video”

  Specifies that the target of the augmentation is a video.

• process_type (str) –

  Type of the augmentation.

  The options are:

  ”batch”

  Batch type means that the input and output paths are directories.

  ”single”

  Single type means that the input and output paths are files.

• mode (str) –

  Mode of the augmentation.

  The options are:

  ”sequential”

  Sequential mode means that the augmentations will be applied one by one. This means that the data will be saved after each augmentation.

  ”singular”

  Singular mode means that the augmentations will be applied all at once. This means that the data will be saved only after all augmentations were applied.

• augmentations (dict) – Dict of augmentations to be applied to the data. The keys are the names of the augmentations and the values are the augmentations themselves.

• verbose (bool, optional) – If True, prints the overall progress of the augmentation process. The default is False.

• aug_verbose (bool, optional) – If True, prints the progress of the augmentation process. The default is False.

• warn_verbose (bool, optional) – If True, prints the warnings. The default is False.

Return type

None.

class CVAugmentor.assets.main.augmentations_class.Augmentations

Class containing the image augmentations.

The augmentations include:

no_augmentation, flip, zoom, rotate, shear, grayscale, hue, saturation, brightness, exposure, blur, noise, cutout, negative.

static blur(radius: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Blur the image. The blur radius is the standard deviation of the Gaussian blur.

Parameters

radius (int or float, optional) – Blur radius. The default is None. If None, a random blur radius is generated.

Returns

blur_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static brightness(brightness_factor: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Adjust the brightness of the image.

Parameters

brightness_factor (int or float, optional) – Brightness factor. The default is None. If None, a random brightness factor is generated.

Returns

brightness_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static cutout(max_count: Optional[int] = None, max_size: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Cutout a random part of the image. This method cuts out a random part of the image and replaces it with a black rectangle.

Parameters

• max_size (int or float, optional) – Maximum size of the cutout. The default is None. If None, a random maximum size is generated.

• max_count (int, optional) – Maximum number of cutouts. The default is None. If None, a random maximum count is generated.

Returns

cutout_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static exposure(exposure_factor: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Adjust the exposure of the image. The exposure means the amount of light that reaches the image sensor.

Parameters

exposure_factor (int or float, optional) – Exposure factor. The default is None. If None, a random exposure factor is generated.

Returns

exposure_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static flip(flip_type: Optional[str] = 'vertical') → Callable[[PIL.Image.Image], PIL.Image.Image]

Flip the image horizontally or vertically.

Parameters

flip_type (str, optional) –

Type of flip. The default is “vertical”.

The options are:

”horizontal”

Flip the image horizontally

”vertical”

Flip the image vertically

Returns

flip_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static grayscale() → Callable[[PIL.Image.Image], PIL.Image.Image]

Convert the image to grayscale. Note that the grayscale conversion is keeping the image in RGB format in order to be consistent with the other methods.

Parameters

None. –

Returns

grayscale_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static hue(hue_shift: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Shift the hue of the image. Hue is the color of the image.

Parameters

hue_shift (int or float, optional) – Hue shift value. The default is None. If None, a random hue shift value is generated.

Returns

hue_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static negative() → Callable[[PIL.Image.Image], PIL.Image.Image]

Convert the image to negative. This method converts the image to negative by inverting the pixel values.

Parameters

None. –

Returns

negative_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static no_augmentation() → Callable[[PIL.Image.Image], PIL.Image.Image]

No augmentation. This can be used when you want to copy the image to the output folder without augmenting it.

Parameters

None. –

Returns

no_augmentation_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static noise(intensity: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Add noise to the image. This method adds noise to the image by multiplying the pixel values by a random number between -1 and 1. It creates a noisy image that looks like the output from an analog TV with bad reception.

Parameters

intensity (int or float, optional) – Noise intensity. The default is None. If None, a random noise intensity is generated.

Returns

noise_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static rotate(rotate_type: Optional[str] = 'definite', angle: Optional[Union[int, float]] = 90) → Callable[[PIL.Image.Image], PIL.Image.Image]

Rotate the image by a definite angle or a random angle.

Parameters

• rotate_type (str, optional) –

  Type of rotation. The default is “definite”.

  The options are:

  ”definite”

  Rotate the image by a definite angle

  ”random”

  Rotate the image by a random angle

• angle (int or float, optional if rotate_type is "random") – Angle of rotation. The default is 90.

Returns

rotate_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static saturation(saturation_factor: Optional[Union[int, float]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Adjust the saturation of the image. The stauration means the intensity of the colors.

Parameters

saturation_factor (int or float, optional) – Saturation factor. The default is None. If None, a random saturation factor is generated.

Returns

saturation_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static shear(shear: Optional[Tuple[Union[int, float], Union[int, float]]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Shear the image. The shearing means that the image is slanted along the x-axis and/or y-axis.

Parameters

shear (tuple, optional) – Shear values. The default is None. If None, a random shear value is generated.

Returns

shear_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function

static zoom(zoom_size: Optional[Tuple[Union[int, float], Union[int, float]]] = None) → Callable[[PIL.Image.Image], PIL.Image.Image]

Zoom the image.

Parameters

zoom_size (tuple, optional) – Size of the zoom. The default is None. If None, a random zoom size is generated.

Returns

zoom_wrapper –

Args:

imageImage.Image

The image to be augmented.

Return type

function