preprocesing.layout_engine package¶

Submodules¶

preprocesing.layout_engine.helpers module¶

preprocesing.layout_engine.helpers.choose(parent)¶

Choose one panel from a parent panel’s children

Parameters: parent – Parent panel whose children are

chosen from

Returns: Index of chosen child panel
Return type: int

preprocesing.layout_engine.helpers.choose_and_return_other(parent)¶

Choose a particular panel to return randomly and also return panels not chosen

Parameters: parent – parent panel whose

children are picked from

Returns: A tuple of a chosen child panel index

and a list the indices of those that weren’t

Return type: tuple

preprocesing.layout_engine.helpers.crop_image_only_outside(img, tol=0)¶

Crop the outside of the image where the pixels are black

Parameters

img (PIL.Image) – image to be cropped
tol (int, optional) – tollerance level, defaults to 0

Returns

Cropped image

Return type

PIL.Image

preprocesing.layout_engine.helpers.find_parent_with_multiple_children(page, n)¶

This function finds parent panels which have a particular number of children

Parameters

page (Page) – Page to be searched
n (int) – number of children

Returns

The children of the page

which hae n children

Return type: list

preprocesing.layout_engine.helpers.get_leaf_panels(page, panels)¶

Get panels which are to actually be rendered recursively i.e. they are the leaves of the Page-Panel tree

Parameters

page (Page) – Page to be searched
panels – A list of panels to be

returned by refernce

preprocesing.layout_engine.helpers.get_min_area_panels(panel, min_area=0.1, ret_panels=[])¶

Recursively get a set of panels which have a particular minimum area

Parameters

panel (Panel) – parent panel
min_area – Minimum area as a ratio

of the page’s area, defaults to 0.1

Parameters: ret_panels (list, optional) – Panels to return, defaults to []

preprocesing.layout_engine.helpers.invert_for_next(current)¶

A helper function used to invert the orientation indicator of the next panel based on the current panel i.e. if a parent is horizontal a child panel must be vertical

Parameters: current – Current parent

orientation

Returns: child panel orientation
Return type: str

preprocesing.layout_engine.helpers.move_child_to_line(point, change, old_line, orientation)¶

A helper function that uses the triangle similarity theorem which specifies that if one triangle is just a sub triangle of another i.e. is just one line drawn from any two sides it’s sides share a ratio.

Parameters

point (tuple) – coordinate to be moved
change – amount the original line was moved

as a length

Parameters

old_line (tuple) – original line which was moved
orientation – orientation of panel whose

line was moved therefore the line

Returns: Amount the point needs to be adjusted
Return type: float

preprocesing.layout_engine.helpers.move_children_to_line(parent, line, change, orientation, direction)¶

A helper function that recursively moves the children of a parent panel to a particular line which is where the parent panel’s new side lines using a basic trignometric formula of similar triangles

Parameters

parent (Panel) – Parent panel which is being transformed
line – A set of xy coordinates to move the

child panels to

Parameters

change (float) – How much the parent panel’s line moved
orientation – orientation of the parent panel

i.e. horizontal or vertical

Parameters: direction (str) – which of the line’s sides went up

preprocesing.layout_engine.page_creator module¶

preprocesing.layout_engine.page_creator.create_single_page(data)¶

This function is used to render a single page from a metadata json file to a target location.

Parameters: paths – a tuple of the page metadata and output path

as well as whether or not to save the rendered file i.e. dry run or wet run

preprocesing.layout_engine.page_creator.render_pages(metadata_dir, images_dir, dry=False)¶

Takes metadata json files and renders page images

Parameters

metadata_dir (str) – A directory containing all the metadata json files
images_dir (str) – The output directory for the rendered pages

preprocesing.layout_engine.page_dataset_creator module¶

preprocesing.layout_engine.page_dataset_creator.add_background(page, image_dir, image_dir_path)¶

Add a background image to the page

Parameters

page (Page) – Page to add background to
image_dir (list) – A list of images
image_dir_path – path to images used for adding

the full path to the page

Returns: Page with background
Return type: Page

preprocesing.layout_engine.page_dataset_creator.add_transforms(page)¶

Adds panel boundary transformations to the page

Parameters: page (Page) – Page to be transformed
Returns: Page with transformed panels
Return type: Page

preprocesing.layout_engine.page_dataset_creator.box_transform_page(page, direction_list=[])¶

This function takes all the first child panels of a page and moves them to form a zigzag or a rhombus pattern

Parameters

page (Page) – Page to be transformed
direction_list – A list of directions the page

should move it’s child panel’s corner’s to

Returns: Transformed page
Return type: Page

preprocesing.layout_engine.page_dataset_creator.box_transform_panels(page, type_choice=None, pattern=None)¶

This function move panel boundaries to transform them into trapezoids and rhombuses

Parameters

page (Page) – Page to be transformed
type_choice – If you want to specify

a particular transform type: rhombus or trapezoid, defaults to None

Parameters: pattern – Based on the type choice choose a pattern.

For rhombus it’s left or right, for trapezoid it’s A or V defaults to None

Returns: Transformed Page
Return type: Page

preprocesing.layout_engine.page_dataset_creator.create_page_metadata(image_dir, image_dir_path, font_files, text_dataset, speech_bubble_files, speech_bubble_tags)¶

This function creates page metadata for a single page. It includes transforms, background addition, random panel removal, panel shrinking, and the populating of panels with images and speech bubbles.

Parameters

image_dir (list) – List of images to pick from
image_dir_path – Path of images dir to add to

panels

Parameters: font_files – list of font files for speech bubble

text

Parameters: text_dataset – A dask dataframe of text to

pick to render within speech bubble

Parameters: speech_bubble_files – list of base speech bubble

template files

Parameters: speech_bubble_tags – a list of speech bubble

writing area tags by filename

Returns: Created Page with all the bells and whistles
Return type: Page

preprocesing.layout_engine.page_dataset_creator.create_single_panel_metadata(panel, image_dir, image_dir_path, font_files, text_dataset, speech_bubble_files, speech_bubble_tags, minimum_speech_bubbles=0)¶

This is a helper function that populates a single panel with an image, and a set of speech bubbles

Parameters

panel (Panel) – Panel to add image and speech bubble to
image_dir (list) – List of images to pick from
image_dir_path – Path of images dir to add to

panels

Parameters: font_files – list of font files for speech bubble

text

Parameters: text_dataset – A dask dataframe of text to

pick to render within speech bubble

Parameters: speech_bubble_files – list of base speech bubble

template files

Parameters: speech_bubble_tags – a list of speech bubble

writing area tags by filename

Parameters: minimum_speech_bubbles – Set whether panels

have a minimum number of speech bubbles, defaults to 0

preprocesing.layout_engine.page_dataset_creator.draw_n(n, parent, horizontal_vertical)¶

A function to take a parent Panel and divide it into n sub-panel’s vertically or horizontally with each panels having equal size ratios along the axis perpendicular to their orientation

NOTE: This function performs actions by reference

Parameters

n (int) – Number of sub-panels
parent (Panel) – The parent panel being split
horizontal_vertical – Whether to render the sub-panels vertically

or horizontally in regards to the page

preprocesing.layout_engine.page_dataset_creator.draw_n_shifted(n, parent, horizontal_vertical, shifts=[])¶

A function to take a parent Panel and divide it into n sub-panel’s vertically or horizontally with each panels having specified size ratios along the axis perpendicular to their orientation

NOTE: This function performs actions by reference

Parameters

n (int) – Number of sub-panels
parent (Panel) – The parent panel being split
horizontal_vertical – Whether to render the sub-panels vertically

or horizontally in regards to the page

Parameters: shifts (list) – Ratios to divide the panel into sub-panels

preprocesing.layout_engine.page_dataset_creator.draw_two_shifted(parent, horizontal_vertical, shift=None)¶

Draw two subpanels of a parent panel

Parameters

parent (Parent) – The parent panel to be split
horizontal_vertical – Orientation of sub-panels in refrence

to the page

Parameters: shift (float, optional) – by what ratio should the 2 panels be split, defaults to None

preprocesing.layout_engine.page_dataset_creator.get_base_panels(num_panels=0, layout_type=None, type_choice=None, page_name=None)¶

This function creates the base panels for one page it specifies how a page should be layed out and how many panels should be in it

Parameters: num_panels – how many panels should be on a page

if 0 then the function chooses, defaults to 0

Parameters: layout_type – whether the page should consist of

vertical, horizontal or both types of panels, defaults to None

Parameters: type_choice – If having selected vh panels select a type

of layout specifically, defaults to None

Parameters: page_name (str, optional) – A specific name for the page
Returns: A Page object with the panels initalized
Return type: Page

preprocesing.layout_engine.page_dataset_creator.populate_panels(page, image_dir, image_dir_path, font_files, text_dataset, speech_bubble_files, speech_bubble_tags, minimum_speech_bubbles=0)¶

This function takes all the panels and adds backgorund images and speech bubbles to them

Parameters

page (Page) – Page with panels to populate
image_dir (list) – List of images to pick from
image_dir_path – Path of images dir to add to

panels

Parameters: font_files – list of font files for speech bubble

text

Parameters: text_dataset – A dask dataframe of text to

pick to render within speech bubble

Parameters: speech_bubble_files – list of base speech bubble

template files

Parameters: speech_bubble_tags – a list of speech bubble

writing area tags by filename

Parameters: minimum_speech_bubbles – Set whether panels

have a minimum number of speech bubbles, defaults to 0

Returns: Page with populated panels
Return type: Page

preprocesing.layout_engine.page_dataset_creator.remove_panel(page)¶

This function randomly removes a panel from pages which have more than n+1 panels

Parameters: page (Page) – Page to remove panels from
Returns: Page with panels removed
Return type: Page

preprocesing.layout_engine.page_dataset_creator.shrink_panels(page)¶

A function that uses the pyclipper library] to reduce the size of the panel polygon

Parameters: page – Page whose panels are to be

shrunk

Returns: Page with shrunk panels
Return type: Page

preprocesing.layout_engine.page_dataset_creator.single_slice_panels(page, horizontal_vertical=None, type_choice=None, skew_side=None, number_to_slice=0)¶

Slices a panel once at an angle into two new panels

Parameters

page (Page) – Page to have panels sliced
horizontal_vertical – Whether the slice should be horizontal

or vertical

Parameters: type_choice – Specify whether the panel should be

sliced down the “center” or on a “side”, defaults to None

Parameters: skew_side – Based on the type of slicing which direction should

it be sliced

Parameters: number_to_slice (int) – Number of panels to slice
Returns: page with sliced panels
Return type: Page

preprocesing.layout_engine.page_object_classes module¶

class preprocesing.layout_engine.page_object_classes.Page(coords=[], page_type='', num_panels=1, children=[], name=None)¶

Bases: preprocesing.layout_engine.page_object_classes.Panel

A class that represents a full page consiting of multiple child panels

Parameters

coords (list) – A list of the boundary coordinates of a page
page_type – Signifies whether a page consists of vertical

or horizontal panels or both

Parameters

num_panels (int) – Number of panels in this page
children (list, optional:) – List of direct child panels of this page

dump_data(dataset_path, dry=True)¶

A method to take all the Page’s relevant data and create a dictionary out of it so it can be exported to JSON so that it can then be loaded and rendered to images in parallel

Parameters

dataset_path (str) – Where to dump the JSON file
dry (bool, optional) – Whether to just return or write the JSON file

Returns

Optional return when running dry of a json data dump

Return type

str

load_data(filename)¶

This method reverses the dump_data function and load’s the metadata of the page from the JSON file that has been loaded.

Parameters: filename (str) – JSON filename to load

render(show=False)¶

A function to render this page to an image

Parameters: show (bool, optional) – Whether to return this image or to show it

class preprocesing.layout_engine.page_object_classes.Panel(coords, name, parent, orientation, children=[], non_rect=False)¶

Bases: object

A class to encapsulate a panel of the manga page. Since the script works in a parent-child relationship where each panel child is an area subset of some parent panel, some panels aren’t leaf nodes and thus not rendered.

Parameters

coords (list) – Coordinates of the boundary of the panel
name (str) – Unique name for the panel
parent (Panel) – The panel which this panel is a child of
orientation – Whether the panel consists of lines that are vertically

or horizotnally oriented in reference to the page

Children: Children panels of this panel
Non_rect: Whether the panel was transformed to be non rectangular

and thus has less or more than 4 coords

add_child(panel)¶

Add child panels

Parameters: panel (Panel) – A child panel to the current panel

add_children(panels)¶

Method to add multiple children at once

Parameters: panels (list) – A list of Panel objects

dump_data()¶

A method to take all the Panel’s relevant data and create a dictionary out of it so it can be exported to JSON via the Page(Panel) class’s dump_data method

Returns: A dictionary of the Panel’s data
Return type: dict

get_child(idx)¶

Get a child panel by index

Parameters: idx (int) – Index of a child panel
Returns: The child at the idx
Return type: Panel

get_polygon()¶

Return the coords in a format that can be used to render a polygon via Pillow

Returns: A tuple of coordinate tuples of the polygon’s vertices
Return type: tuple

load_data(data)¶

This method reverses the dump_data function and load’s the metadata of the panel from the subsection of the JSON file that has been loaded

Parameters: data (dict) – A dictionary of this panel’s data

refresh_coords()¶: When chances are made to the xy coordinates variables directly this function allows you to refresh the coords variable with the changes

refresh_vars()¶: When chances are made to the xy coordinates directly this function allows you to refresh the x1y1… variable with the changes

class preprocesing.layout_engine.page_object_classes.SpeechBubble(texts, text_indices, font, speech_bubble, writing_areas, resize_to, location, width, height, transforms=None, transform_metadata=None, text_orientation=None)¶

Bases: object

A class to represent the metadata to render a speech bubble

Parameters: texts – A list of texts from the text corpus to render in this

bubble

Parameters: text_indices – The indices of the text from the dataframe

for easy retrival

Parameters

font (str) – The path to the font used in the bubble
speech_bubble – The path to the base speech bubble file

used for this bubble

Parameters: writing_areas – The areas within the bubble where it is okay

to render text

Parameters: resize_to – The amount of area this text bubble should consist of

which is a ratio of the panel’s area

Parameters: location – The location of the top left corner of the speech bubble

on the page

Parameters

width (float) – Width of the speech bubble
height (float) – Height of the speech bubble
transforms – A list of transformations to change

the shape of the speech bubble

Parameters: transform_metadata – Metadata associated with transformations,

defaults to None

Parameters: text_orientation – Whether the text of this speech bubble

is written left to right ot top to bottom

dump_data()¶

A method to take all the SpeechBubble’s relevant data and create a dictionary out of it so it can be exported to JSON via the Page(Panel) class’s dump_data method

Returns: Data to be returned to Page(Panel) class’s

dump_data method :rtype: dict

render()¶

A function to render this speech bubble

Returns: A list of states of the speech bubble,

the speech bubble itself, it’s mask and it’s location on the page :rtype: tuple

preprocesing.layout_engine package¶

Submodules¶

preprocesing.layout_engine.helpers module¶

preprocesing.layout_engine.page_creator module¶

preprocesing.layout_engine.page_dataset_creator module¶

preprocesing.layout_engine.page_object_classes module¶

Module contents¶