Source: vis/utils/utils.py#L0

Global Variables

  • slicer

reverse_enumerate

reverse_enumerate(iterable)

Enumerate over an iterable in reverse order while retaining proper indexes, without creating any copies.

listify

listify(value)

Ensures that the value is a list. If it is not a list, it creates a new list with value as an item.

add_defaults_to_kwargs

add_defaults_to_kwargs(defaults, **kwargs)

Updates kwargs with dict of defaults

Args:

  • defaults: A dictionary of keys and values **kwargs: The kwargs to update.

Returns:

The updated kwargs.

get_identifier

get_identifier(identifier, module_globals, module_name)

Helper utility to retrieve the callable function associated with a string identifier.

Args:

  • identifier: The identifier. Could be a string or function.
  • module_globals: The global objects of the module.
  • module_name: The module name

Returns:

The callable associated with the identifier.

apply_modifications

apply_modifications(model, custom_objects=None)

Applies modifications to the model layers to create a new Graph. For example, simply changing model.layers[idx].activation = new activation does not change the graph. The entire graph needs to be updated with modified inbound and outbound tensors because of change in layer building function.

Args:

  • model: The keras.models.Model instance.

Returns:

The modified model with changes applied. Does not mutate the original model.

random_array

random_array(shape, mean=128.0, std=20.0)

Creates a uniformly distributed random array with the given mean and std.

Args:

  • shape: The desired shape
  • mean: The desired mean (Default value = 128)
  • std: The desired std (Default value = 20)

Returns: Random numpy array of given shape uniformly distributed with desired mean and std.

find_layer_idx

find_layer_idx(model, layer_name)

Looks up the layer index corresponding to layer_name from model.

Args:

  • model: The keras.models.Model instance.
  • layer_name: The name of the layer to lookup.

Returns:

The layer index if found. Raises an exception otherwise.

deprocess_input

deprocess_input(input_array, input_range=(0, 255))

Utility function to scale the input_array to input_range throwing away high frequency artifacts.

Args:

  • input_array: An N-dim numpy array.
  • input_range: Specifies the input range as a (min, max) tuple to rescale the input_array.

Returns:

The rescaled input_array.

stitch_images

stitch_images(images, margin=5, cols=5)

Utility function to stitch images together with a margin.

Args:

  • images: The array of 2D images to stitch.
  • margin: The black border margin size between images (Default value = 5)
  • cols: Max number of image cols. New row is created when number of images exceed the column size. (Default value = 5)

Returns:

A single numpy image array comprising of input images.

get_img_shape

get_img_shape(img)

Returns image shape in a backend agnostic manner.

Args:

  • img: An image tensor of shape: (channels, image_dims...) if data_format='channels_first' or (image_dims..., channels) if data_format='channels_last'.

Returns:

Tuple containing image shape information in (samples, channels, image_dims...) order.

load_img

load_img(path, grayscale=False, target_size=None)

Utility function to load an image from disk.

Args:

  • path: The image file path.
  • grayscale: True to convert to grayscale image (Default value = False)
  • target_size: (w, h) to resize. (Default value = None)

Returns:

The loaded numpy image.

lookup_imagenet_labels

lookup_imagenet_labels(indices)

Utility function to return the image net label for the final dense layer output index.

Args:

  • indices: Could be a single value or an array of indices whose labels should be looked up.

Returns:

Image net label corresponding to the image category.

draw_text

draw_text(img, text, position=(10, 10), font="FreeSans.ttf", font_size=14, color=(0, 0, 0))

Draws text over the image. Requires PIL.

Args:

  • img: The image to use.
  • text: The text string to overlay.
  • position: The text (x, y) position. (Default value = (10, 10))
  • font: The ttf or open type font to use. (Default value = 'FreeSans.ttf')
  • font_size: The text font size. (Default value = 12)
  • color: The (r, g, b) values for text color. (Default value = (0, 0, 0))

Returns: Image overlayed with text.

bgr2rgb

bgr2rgb(img)

Converts an RGB image to BGR and vice versa

Args:

  • img: Numpy array in RGB or BGR format

Returns: The converted image format

normalize

normalize(array, min_value=0.0, max_value=1.0)

Normalizes the numpy array to (min_value, max_value)

Args:

  • array: The numpy array
  • min_value: The min value in normalized array (Default value = 0)
  • max_value: The max value in normalized array (Default value = 1)

Returns:

The array normalized to range between (min_value, max_value)