Image Generators

AudiverisOmrImageGenerator Module

class omrdatasettools.AudiverisOmrImageGenerator.AudiverisOmrImageGenerator

AudiverisOmrImageGenerator.extract_symbols(raw_data_directory: str, destination_directory: str)

Extracts the symbols from the raw XML documents and matching images of the Audiveris OMR dataset into individual symbols

Parameters:

• raw_data_directory – The directory, that contains the xml-files and matching images

• destination_directory – The directory, in which the symbols should be generated into. One sub-folder per symbol category will be generated automatically

CapitanImageGenerator Module

class omrdatasettools.CapitanImageGenerator.CapitanImageGenerator

CapitanImageGenerator.create_capitan_images(raw_data_directory: str, destination_directory: str, stroke_thicknesses: List[int]) → None

Creates a visual representation of the Capitan strokes by parsing all text-files and the symbols as specified by the parameters by drawing lines that connect the points from each stroke of each symbol.

Parameters:

• raw_data_directory – The directory, that contains the raw capitan dataset

• destination_directory – The directory, in which the symbols should be generated into. One sub-folder per symbol category will be generated automatically

• stroke_thicknesses – The thickness of the pen, used for drawing the lines in pixels. If multiple are specified, multiple images will be generated that have a different suffix, e.g. 1-16-3.png for the 3-px version and 1-16-2.png for the 2-px version of the image 1-16

HomusImageGenerator Module

class omrdatasettools.HomusImageGenerator.HomusImageGenerator

static HomusImageGenerator.create_images(raw_data_directory: str, destination_directory: str, stroke_thicknesses: List[int], canvas_width: int = None, canvas_height: int = None, staff_line_spacing: int = 14, staff_line_vertical_offsets: List[int] = None, random_position_on_canvas: bool = False) → dict

Creates a visual representation of the Homus Dataset by parsing all text-files and the symbols as specified by the parameters by drawing lines that connect the points from each stroke of each symbol.

Each symbol will be drawn in the center of a fixed canvas, specified by width and height.

Parameters:

• raw_data_directory – The directory, that contains the text-files that contain the textual representation of the music symbols

• destination_directory – The directory, in which the symbols should be generated into. One sub-folder per symbol category will be generated automatically

• stroke_thicknesses – The thickness of the pen, used for drawing the lines in pixels. If multiple are specified, multiple images will be generated that have a different suffix, e.g. 1-16-3.png for the 3-px version and 1-16-2.png for the 2-px version of the image 1-16

• canvas_width – The width of the canvas, that each image will be drawn upon, regardless of the original size of the symbol. Larger symbols will be cropped. If the original size of the symbol should be used, provided None here.

• canvas_height – The height of the canvas, that each image will be drawn upon, regardless of the original size of the symbol. Larger symbols will be cropped. If the original size of the symbol should be used, provided None here

• staff_line_spacing – Number of pixels spacing between each of the five staff-lines

• staff_line_vertical_offsets – List of vertical offsets, where the staff-lines will be superimposed over the drawn images. If None is provided, no staff-lines will be superimposed. If multiple values are provided, multiple versions of each symbol will be generated with the appropriate staff-lines, e.g. 1-5_3_offset_70.png and 1-5_3_offset_77.png for two versions of the symbol 1-5 with stroke thickness 3 and staff-line offsets 70 and 77 pixels from the top.

• random_position_on_canvas – True, if the symbols should be randomly placed on the fixed canvas. False, if the symbols should be centered in the fixed canvas. Note that this flag only has an effect, if fixed canvas sizes are used.

Returns:

A dictionary that contains the file-names of all generated symbols and the respective bounding-boxes of each symbol.

static HomusImageGenerator.add_arguments_for_homus_image_generator(parser: ArgumentParser)

HomusSymbol Module

class omrdatasettools.HomusImageGenerator.HomusSymbol(content: str, strokes: List[List[Point2D]], symbol_class: str, dimensions: Rectangle)

static HomusSymbol.initialize_from_string(content: str) → HomusSymbol

Create and initializes a new symbol from a string

Parameters:

content – The content of a symbol as read from the text-file

Returns:

The initialized symbol

Return type:

HomusSymbol

HomusSymbol.draw_into_bitmap(export_path: ExportPath, stroke_thickness: int, margin: int = 0) → None

Draws the symbol in the original size that it has plus an optional margin

Parameters:

• export_path – The path, where the symbols should be created on disk

• stroke_thickness – Pen-thickness for drawing the symbol in pixels

• margin – An optional margin for each symbol

HomusSymbol.draw_onto_canvas(export_path: ExportPath, stroke_thickness: int, margin: int, destination_width: int, destination_height: int, staff_line_spacing: int = 14, staff_line_vertical_offsets: List[int] = None, bounding_boxes: dict = None, random_position_on_canvas: bool = False) → None

Draws the symbol onto a canvas with a fixed size

Parameters:

• bounding_boxes – The dictionary into which the bounding-boxes will be added of each generated image

• export_path – The path, where the symbols should be created on disk

• stroke_thickness –

• margin –

• destination_width –

• destination_height –

• staff_line_spacing –

• staff_line_vertical_offsets – Offsets used for drawing staff-lines. If None provided, no staff-lines will be drawn if multiple integers are provided, multiple images will be generated

MeasureVisualizer Module

This class can be used to generate visualizations of measure annotations, such as this one for the Muscima++ dataset:

class omrdatasettools.MeasureVisualizer.MeasureVisualizer(draw_system_measures: bool, draw_stave_measures: bool, draw_staves: bool)

Class that can be used to visualize the measure annotations that are provided as json files. Allows to enable/disable whether to draw system-measures (one measure for all instruments), stave-measures (one measure, single stave of one instrument) and staves (the whole line of a single instrument). System-measures contain all stave-measures that are being played simulatenously.

Bounding boxes will be drawn with semitransparent colors.

MeasureVisualizer.draw_bounding_boxes_for_all_images_in_directory(image_directory, json_annotation_directory)

MeasureVisualizer.draw_bounding_boxes_into_image(image_path: str, ground_truth_annotations_path: str)

MuscimaPlusPlusMaskImageGenerator Module

class omrdatasettools.MuscimaPlusPlusMaskImageGenerator.MaskType(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)

The type of masks that should be generated

NODES_SEMANTIC_SEGMENTATION = 1

creates mask images, where each type of node gets the same color mask (for semantic segmentation). The classes staffLine, staff and staffSpace are ignored.

STAFF_BLOBS_INSTANCE_SEGMENTATION = 3

creates mask images, where each staff will receive one big blob (filling the staff space regions) per staff line for instance segmentation. So each staff will have a different color.

STAFF_LINES_INSTANCE_SEGMENTATION = 2

creates mask images, where the masks of the staff lines are contained for instance segmentation. All five lines that form a staff will have the same color.

class omrdatasettools.MuscimaPlusPlusMaskImageGenerator.MuscimaPlusPlusMaskImageGenerator

MuscimaPlusPlusMaskImageGenerator.render_node_masks(raw_data_directory: str, destination_directory: str, mask_type: MaskType)

Extracts all symbols from the raw XML documents and generates individual symbols from the masks

Parameters:

• raw_data_directory – The directory, that contains the xml-files and matching images

• destination_directory – The directory, in which the symbols should be generated into. Per file, one mask will be generated.

• mask_type – The type of masks that you want to generate, e.g., masks for each node or staff lines only.

MuscimaPlusPlusSymbolImageGenerator Module

class omrdatasettools.MuscimaPlusPlusSymbolImageGenerator.MuscimaPlusPlusSymbolImageGenerator

MuscimaPlusPlusSymbolImageGenerator.extract_and_render_all_symbol_masks(raw_data_directory: str, destination_directory: str)

Extracts all symbols from the raw XML documents and generates individual symbols from the masks

Parameters:

• raw_data_directory – The directory, that contains the xml-files and matching images

• destination_directory – The directory, in which the symbols should be generated into. One sub-folder per symbol category will be generated automatically