infy-bordered-table-extractor.md

1. Introduction

infy_bordered_table_extractor(v0.0.7) is a python package that can extract data from an image of a bordered table. The output can be a dictionary or an excel file.

The library requires python 3.6

2. Build and Install

For build:

python setup.py bdist_wheel

For install:

pipenv install <whl file>

3. API

Initialization

Creates an instance of Bordered Table Extractor.

class BorderedTableExtractor(table_detection_provider:infy_bordered_table_extractor.interface.data_service_provider_interface.DataServiceProviderInterface,
		cell_extraction_provider:infy_bordered_table_extractor.interface.data_service_provider_interface.DataServiceProviderInterface,
		temp_folderpath:str,
	logger:logging.Logger=None,
	debug_mode_check:bool=False):

Input:

Argument	Description
table_detection_provider (DataServiceProviderInterface)	Provider for table detection
cell_extraction_provider (DataServiceProviderInterface)	Provider for cell data extraction
temp_folderpath (str)	Path to temp folder
logger (logging.Logger, optional)	Path to store all the debug info files. Defaults to None.
debug_mode_check (bool, optional)	To get debug info while using the API. Defaults to False.

Output:

None

API - extract_all_fields

Extracts table information from an image and returns as an json output and saves it as an excel file, if required.

def extract_all_fields(image_file:str,
		file_data_list:[{'path': '',
	'pages': []}]=None,
	within_bbox:list=None,
		config_param_dict:{'custom_cells': [{'rows': [],
		'columns': []}],
		'col_header': {'use_first_row': True,
		'values': []},
		'deskew_image_reqd': False,
		'image_cell_cleanup_reqd': True,
		'output': {'path': None,
		'format': None},
		'rgb_line_skew_detection_method': [<RgbSkewDetectionMethod.CONVOLUTION_CONTRAST_METHOD: 1>],
	'line_detection_method': [<LineDetectionMethod.RGB_LINE_DETECT: 1>]}=None):

Input:

Argument	Description
image_file (str)	image file path for which table has to be extracted
file_data_list ([FILE_DATA], optional)	List of all supporting file paths and page numbers, if applicable. Image and the file must have the same content.
within_bbox (list, optional)	Bounding box coordinates(x, y, width, height) of the table in the image. Defaults to None.
config_param_dict (CONFIG_PARAM_DICT, optional)	Additional info. Defaults to None. custom_cells: customized cell extraction col_header: Choose the header to be used use_first_row. To customize values for header use values. deskew_image_reqd: If detected skew in image is to be corrected image_cell_cleanup_reqd: If each cell image is required for cleaning for text extraction output: path to specify where to save the file and format for type of file to save the data rgb_line_skew_detection_method: List of all skew detection methods for detecting skewness line_detection_method: list of all line detection methods for line detection

Output:

Param	Description
dict	Dict of saved info.

Initialization

Creates an instance of Tesseract Data Service Provider

class TesseractDataServiceProvider(tesseract_path:str,
	logger:logging.Logger=None,
	log_level:int=None):

Input:

Argument	Description
tesseract_path (str)	Path to Tesseract.
logger (logging.Logger, optional)	Logger object. Defaults to None.
log_level (int, optional)	Logging Level. Defaults to None.

Output:

None

API - get_text

Method to be implemented to return the text from the list of cell images or bbox of the original image as a list of dictionary. (Eg. [{'cell_id': str,'cell_text':'{{extracted_text}}', 'cell_bbox':[x, y, w, h]}]

def get_text(img:<built-in function array>,
		img_cell_bbox_list:[{'cell_id': '',
		'cell_bbox': []}],
		file_data_list:[{'path': '',
	'pages': []}]=None,
		additional_info:{'cell_info': [{'cell_id': '',
		'cell_img_path': '',
		'cell_bbox': [],
	'cell_img': <built-in function array>}]}=None,
	temp_folderpath:str=None) -> [{'cell_id': '',
		'cell_text': '',
		'cell_bbox': []}]:

Input:

Argument	Description
img (np.array)	Read image as np array of the original image.
img_cell_bbox_list ([IMG_CELL_BBOX])	List of all cell bbox
file_data_list (FILE_DATA,optional)	List of all file datas. Each file data has the path to supporting document and page numbers, if applicable. Defaults to None.
additional_info (ADDITIONAL_INFO, optional)	Additional info. Defaults to None.
temp_folderpath (str, optional)	Path to temp folder. Defaults to None.

Output:

Param	Description
[GET_TEXT_OUTPUT]	list of dict containing text and its bbox.

API - get_tokens

Method to be implemented to get all tokens (word, phrase or line) and its bounding box as x, y, width and height from an image as a list of dictionary. Currently word token is only required.

def get_tokens(token_type_value:int,
		img:<built-in function array>,
		file_data_list:[{'path': '',
	'pages': []}]=None) -> [{'text': '',
		'bbox': [],
		'conf': <class 'int'>}]:

Input:

Argument	Description
token_type_value (int)	1(WORD), 2(LINE), 3(PHRASE)
img (np.array)	Read image as np array of the original image.
file_data_list ([FILE_DATA], optional)	List of all file datas. Each file data has the path to supporting document and page numbers, if applicable. When multiple files are passed, provider has to pick the right file based on the image dimensions or type of file extension. Defaults to None.

Output:

Param	Description
[GET_TOKENS_OUTPUT]	list of dict containing text and its bbox.