Module fontai.config.preprocessing

Classes

class Config (**data: Any)

Configuration class for the image extraction pipeline stage

Args

output_record_class : records.TfrWritable

tfr-compatible output classes from the module fontai.io.records; currently supported are LabeledChar and LabeledFont

output_array_size : int

size of the final grayscale image corresponding to each font's characters

font_to_array_config : int

Data object with runtime parameters for exctracting image arrays from files

beam_cmd_line_args : t.List[str]

List of command line arguments passed to the Beam pipeline

Create a new model by parsing and validating input data from keyword arguments.

Raises ValidationError if the input data cannot be parsed to form a valid model.

Expand source code

class Config(BasePipelineTransformConfig):
  """
  Configuration class for the image extraction pipeline stage
  
  Args:
      output_record_class (records.TfrWritable): tfr-compatible output classes from the module `fontai.io.records`; currently supported are `LabeledChar` and `LabeledFont`
      output_array_size (int): size of the final grayscale image corresponding to each font's characters
      font_to_array_config (int): Data object with runtime parameters for exctracting image arrays from files
      beam_cmd_line_args (t.List[str]): List of command line arguments passed to the Beam pipeline

  """
  output_record_class: type
  output_array_size: PositiveInt
  max_output_file_size: PositiveFloat
  font_to_array_config: FontExtractionConfig
  beam_cmd_line_args: t.List[str]

  @validator("output_record_class")
  def validate_output_schema(schema_class):
    """Validate input record class
    
    Args:
        schema_class (type)
    
    Returns:
        type: Validated record class
    
    Raises:
        TypeError: If record class not in allowed set
    """
    supported = [records.LabeledChar, records.LabeledFont]
    if schema_class in supported:
      return schema_class
    else:
      raise TypeError(f"supported output_record_classes are {[x.__name__ for x in supported]}")

Ancestors

• BasePipelineTransformConfig
• pydantic.main.BaseModel
• pydantic.utils.Representation

Class variables

var beam_cmd_line_args : List[str]

var font_to_array_config : FontExtractionConfig

var max_output_file_size : pydantic.types.PositiveFloat

var output_array_size : pydantic.types.PositiveInt

var output_record_class : type

Static methods

def validate_output_schema()

Validate input record class

Args

schema_class (type)

Returns

type

Validated record class

Raises

TypeError

If record class not in allowed set

Expand source code

@validator("output_record_class")
def validate_output_schema(schema_class):
  """Validate input record class
  
  Args:
      schema_class (type)
  
  Returns:
      type: Validated record class
  
  Raises:
      TypeError: If record class not in allowed set
  """
  supported = [records.LabeledChar, records.LabeledFont]
  if schema_class in supported:
    return schema_class
  else:
    raise TypeError(f"supported output_record_classes are {[x.__name__ for x in supported]}")

class ConfigHandler

Wrapper for image processing stage's configuration handling logic

Expand source code

class ConfigHandler(BaseConfigHandler):
  """
  Wrapper for image processing stage's configuration handling logic

  """

  @classmethod
  def get_config_schema(self):
    """
    YAML configuration schema:

    output_record_class: name of record class that will populate output files; it has to inherit from `fontai.io.records.TfrWritable` and at the moment only `LabeledChar` and `LabeledFont` are supported. If `LabeledChar`, individual character images are preprocessed and stored in no particular order and saved as PNGs. If `LabeledFont`, characters from a single font are stored together and saved as tensors.
    input_path: Input folder with zipped font files
    output_path: output folder where output Tensorflow records files are stored
    output_array_size: Size of square output images
    max_output_file_size: maximum single-file size for output files in MB
    font_extraction_size: Font size to use when extracting characters to images
    canvas_size: initial image canvas size in which font chars are extracted
    canvas_padding: padding for the initial image canvas to allow for unusally large or convoluted fonts
    beam_cmd_line_args: list of Apache Beam's command line aeguments

    """
    schema = yml.Map({
      "output_record_class": yml.Str(), 
      "output_array_size": yml.Int(),
      "font_extraction_size": yml.Int(), 
      "canvas_size": yml.Int(), 
      "canvas_padding": yml.Int(),
      yml.Optional("input_path", default = None): self.IO_CONFIG_SCHEMA, 
      yml.Optional("output_path", default = None): self.IO_CONFIG_SCHEMA,
      yml.Optional("max_output_file_size", default = 64.0): yml.Float(),
      yml.Optional("beam_cmd_line_args", default = ["--runner", "DirectRunner"]): yml.Seq(yml.Str())
       })

    return schema

  def instantiate_config(self, config: yml.YAML) -> Config:
    """
    Processes a YAML instance to produce an Config instance.

    Args:
        config: YAML object from the strictyaml library

    """
    
    input_path, output_path = config.get("input_path").text, config.get("output_path").text
    output_record_class = getattr(records, config.get("output_record_class").text)
    beam_cmd_line_args = config.data["beam_cmd_line_args"]
    output_array_size = config.get("output_array_size").data
    max_output_file_size = config.get("max_output_file_size").data
    f2a_config = FontExtractionConfig(
      charset = string.ascii_letters + string.digits,
      font_extraction_size = config.get("font_extraction_size").data,
      canvas_size = config.get("canvas_size").data,
      canvas_padding = config.get("canvas_padding").data)


    if f2a_config.canvas_padding >=  f2a_config.canvas_size/2:
      raise ValueError(f"canvas padding value ({f2a_config.canvas_padding}) is too large for canvas size ({f2a_config.canvas_size})")


    logger.info(f"Setting output schema as {output_record_class.__name__}")
    
    return Config(
      output_record_class = output_record_class,
      input_path = input_path, 
      output_path = output_path, 
      output_array_size = output_array_size,
      max_output_file_size = max_output_file_size,
      font_to_array_config = f2a_config,
      beam_cmd_line_args = beam_cmd_line_args,
      yaml = config)

Ancestors

• BaseConfigHandler
• abc.ABC

Static methods

def get_config_schema()

YAML configuration schema:

output_record_class: name of record class that will populate output files; it has to inherit from TfrWritable and at the moment only LabeledChar and LabeledFont are supported. If LabeledChar, individual character images are preprocessed and stored in no particular order and saved as PNGs. If LabeledFont, characters from a single font are stored together and saved as tensors. input_path: Input folder with zipped font files output_path: output folder where output Tensorflow records files are stored output_array_size: Size of square output images max_output_file_size: maximum single-file size for output files in MB font_extraction_size: Font size to use when extracting characters to images canvas_size: initial image canvas size in which font chars are extracted canvas_padding: padding for the initial image canvas to allow for unusally large or convoluted fonts beam_cmd_line_args: list of Apache Beam's command line aeguments

Expand source code

@classmethod
def get_config_schema(self):
  """
  YAML configuration schema:

  output_record_class: name of record class that will populate output files; it has to inherit from `fontai.io.records.TfrWritable` and at the moment only `LabeledChar` and `LabeledFont` are supported. If `LabeledChar`, individual character images are preprocessed and stored in no particular order and saved as PNGs. If `LabeledFont`, characters from a single font are stored together and saved as tensors.
  input_path: Input folder with zipped font files
  output_path: output folder where output Tensorflow records files are stored
  output_array_size: Size of square output images
  max_output_file_size: maximum single-file size for output files in MB
  font_extraction_size: Font size to use when extracting characters to images
  canvas_size: initial image canvas size in which font chars are extracted
  canvas_padding: padding for the initial image canvas to allow for unusally large or convoluted fonts
  beam_cmd_line_args: list of Apache Beam's command line aeguments

  """
  schema = yml.Map({
    "output_record_class": yml.Str(), 
    "output_array_size": yml.Int(),
    "font_extraction_size": yml.Int(), 
    "canvas_size": yml.Int(), 
    "canvas_padding": yml.Int(),
    yml.Optional("input_path", default = None): self.IO_CONFIG_SCHEMA, 
    yml.Optional("output_path", default = None): self.IO_CONFIG_SCHEMA,
    yml.Optional("max_output_file_size", default = 64.0): yml.Float(),
    yml.Optional("beam_cmd_line_args", default = ["--runner", "DirectRunner"]): yml.Seq(yml.Str())
     })

  return schema

Methods

def instantiate_config(self, config: strictyaml.representation.YAML) ‑> Config

Processes a YAML instance to produce an Config instance.

Args

config

YAML object from the strictyaml library

Expand source code

def instantiate_config(self, config: yml.YAML) -> Config:
  """
  Processes a YAML instance to produce an Config instance.

  Args:
      config: YAML object from the strictyaml library

  """
  
  input_path, output_path = config.get("input_path").text, config.get("output_path").text
  output_record_class = getattr(records, config.get("output_record_class").text)
  beam_cmd_line_args = config.data["beam_cmd_line_args"]
  output_array_size = config.get("output_array_size").data
  max_output_file_size = config.get("max_output_file_size").data
  f2a_config = FontExtractionConfig(
    charset = string.ascii_letters + string.digits,
    font_extraction_size = config.get("font_extraction_size").data,
    canvas_size = config.get("canvas_size").data,
    canvas_padding = config.get("canvas_padding").data)


  if f2a_config.canvas_padding >=  f2a_config.canvas_size/2:
    raise ValueError(f"canvas padding value ({f2a_config.canvas_padding}) is too large for canvas size ({f2a_config.canvas_size})")


  logger.info(f"Setting output schema as {output_record_class.__name__}")
  
  return Config(
    output_record_class = output_record_class,
    input_path = input_path, 
    output_path = output_path, 
    output_array_size = output_array_size,
    max_output_file_size = max_output_file_size,
    font_to_array_config = f2a_config,
    beam_cmd_line_args = beam_cmd_line_args,
    yaml = config)

Inherited members

• BaseConfigHandler:
  • from_file
  • from_string

class FontExtractionConfig (**data: Any)

Data class that holds the runtime parameters to extract image arrays from files

Args

charset : str

string containing characters to be extracted from font files

font_extraction_size : int

Font size to use when conveting fonts to images

canvas_size : int

Height and width of buffer array in which fonts will be extracted before being processed further

canvas_padding : int

Padding used in the canvas array when extracting the fonts

Create a new model by parsing and validating input data from keyword arguments.

Raises ValidationError if the input data cannot be parsed to form a valid model.

Expand source code

class FontExtractionConfig(BaseModel):
  """
  Data class that holds the runtime parameters to extract image arrays from files
  
  Args:
      charset (str): string containing characters to be extracted from font files
      font_extraction_size (int): Font size to use when conveting fonts to images
      canvas_size (int): Height and width of buffer array in which fonts will be extracted before being processed further
      canvas_padding (int): Padding used in the canvas array when extracting the fonts
  
  """

  charset: str
  font_extraction_size: PositiveInt
  canvas_size: PositiveInt
  canvas_padding: PositiveInt

Ancestors

• pydantic.main.BaseModel
• pydantic.utils.Representation

Class variables

var canvas_padding : pydantic.types.PositiveInt

var canvas_size : pydantic.types.PositiveInt

var charset : str

var font_extraction_size : pydantic.types.PositiveInt