Image Transforms

Normalize

Normalize an image (works with grayscale, RGB and RGBA images).

Python Usage

from deeplodocus.app.transforms.images import normalize_image

np.array = normalize_image(image, mean=, standard_deviation, mean)

Arguments

• mean: (int, float) The mean pixel value of the output image (if not given, automatically computed from the image).
• standard_deviation: (int, float) The standard deviation of pixel values in the output image (if not given, automatically computed ofr the image).

Deeplodocus Configuration

# Example of usage of normalize_image on a RGB image
name: normalize_image
module: deeplodocus.app.transforms.images
kwargs:
  mean: [127.5, 127.5, 127.5]
  standard_deviation: 255

Resize

Resize the given image to the given shape, using the prescribed method.

Python Usage

from deeplodocus.app.transforms.images import resize

np.array = resize(image, shape, keep_aspect=False, padding=0, method=None)

Arguments

• shape: (Tuple, List) The output width and height of the image respectively.
• keep_aspect: (bool=False) Whether or not to keep the aspect ratio of the image.
• padding: (int=0) The padding to apply if keep_aspect is True.
• method: (str=None) The specfic resizing method to. Can be one of : "nearest", "linear", "cubic". If methid is None, "linear" will be used when downsampling and "cublic" will be used with upsampling.

Deeplodocus Configuration

# Example of including resize

name: resize
module : deeplodocus.app.transforms.images
kwargs:
  shape: [256, 128]
  keep_aspect: True
  padding: 0
  method: nearest

Generic Transforms

TODO