Alternat Reference¶

The alternat reference guide will walk through the implementation for both Generation and Collection, the alternat library, and the app.

Generation - Analyzer¶

Google Analyzer

class alternat.generation.google.analyze.AnalyzeImage[source]

Bases: alternat.generation.base.analyzer.AnalyzeImageBase

Google Analyzer driver class.

Parameters: AnalyzeImageBase ([type]) – Driver base class.

describe_image(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Describe image (used for captioning) - Not availble in Google Computer Vision API

Parameters: image (PIL_IMAGE) – [description]

extract_labels(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Extract labels of image using Google Computer Vision API.

Parameters: image (PIL_IMAGE) – PIL Image object.
Raises: Exception – Google Cloud specific error messages based on request.

handle(image_path: str = None, base64_image: str = None, actions: list = None) → dict[source]

Entry point for the driver. Implements all the action and generates data for rule engine.

Parameters

image_path (str, optional) – Path to image on disk, defaults to None
base64_image (str, optional) – Base64 image string, defaults to None
actions (list, optional) – list of actions to run, defaults to None (all actions execute)

Returns

[description]

Return type

dict

ocr_analysis(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Does OCR analysis using Google Computer Vision API. Also runs the alternat clustering rule if app is configured for it.

Parameters: image (PIL_IMAGE) – PIL Image object.

set_environment_variables()[source]: Sets environment variable GOOGLE_APPLICATION_CREDENTIALS based on config.

Azure Analyzer

class alternat.generation.microsoft.analyze.AnalyzeImage[source]

Bases: alternat.generation.base.analyzer.AnalyzeImageBase

Azure / Microsoft Analyzer driver class.

Parameters: AnalyzeImageBase ([type]) – Driver base class.

describe_image(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Describe image using Azure Vision API.

Parameters: image (PIL_Image) – PIL Image object

extract_labels(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Extract labels of image using Azure Vision API.

Parameters: image (PIL_Image) – PIL Image object.

handle(image_path: str = None, base64_image: str = None, actions: list = None) → dict[source]

Entry point for the driver. Implements all the action and generates data for rule engine.

Parameters

image_path (str, optional) – Path to image on disk, defaults to None
base64_image (str, optional) – Base64 image string, defaults to None
actions (list, optional) – list of actions to run, defaults to None (all actions execute)

Returns

[description]

Return type

dict

is_clean(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>) → bool[source]

Check if the image has proper resolution, and is clean.

:param image:PIL Image object. :type image: PIL_Image :return: [description] :rtype: bool

modifyBoundingBoxData(bounding_box: list)[source]

Transform bounding box data as per the convention. Azure API return bounding box info in the format [left, top, right, top, right, bottom, left, bottom] which is transformed to format [{x: left, y: top}, {x: right, y: top}, {x: right, y: bottom}, {x: left, y: bototm}].

Parameters: bounding_box (list) – Bounding box data form Azure API.
Returns: [description]
Return type: [type]

ocr_analysis(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Does OCR Analysis using Azure Vision API.

Parameters: image (PIL_Image) – PIL Image object.

resize_image(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Resize image (maintaining aspect ratio) if width / height > 5000 pixels (API constrain from Azure)

Parameters: image (PIL_Image) – [description]

Opensource Analyzer

class alternat.generation.opensource.analyze.AnalyzeImage[source]

Bases: alternat.generation.base.analyzer.AnalyzeImageBase

Opensource driver class.

Parameters: AnalyzeImageBase ([type]) – Driver base class.

describe_image(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Describe image using open source solution. Not implemented right now.

Parameters: image (PIL_Image) – PIL Image object

extract_labels(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Extract labels of image using open source solution. Not implemented righ now.

Parameters: image (PIL_Image) – PIL Image object.

handle(image_path: str = None, base64_image: str = None, actions: list = None) → dict[source]

Entry point for the driver. Implements all the action and generates data for rule engine.

Parameters

image_path (str, optional) – Path to image on disk, defaults to None
base64_image (str, optional) – Base64 image string, defaults to None
actions (list, optional) – list of actions to run, defaults to None (all actions execute)

Returns

[description]

Return type

dict

modifyBoundingBoxData(bounding_box: list)[source]

Transform bounding box data as per the convention. EasyOCR return bounding box info in the format [left, top, right, top, right, bottom, left, bottom] which is transformed to format [{x: left, y: top}, {x: right, y: top}, {x: right, y: bottom}, {x: left, y: bototm}].

Parameters: bounding_box (list) – Bounding box data form EasyOCR.
Returns: [description]
Return type: [type]

ocr_analysis(image: <module 'PIL.Image' from '/home/docs/checkouts/readthedocs.org/user_builds/alternat/envs/v0.1.4/lib/python3.8/site-packages/PIL/Image.py'>)[source]

Does OCR Analysis using EasyOCR.

Parameters: image (PIL_Image) – PIL Image object.

Generation - Rule Engine¶

OCR Rule

class alternat.generation.rules.ocr_handler.OCRDataHandler(input_data, confidence_threshold: float = None, ocr_filter_threshold: float = None)[source]

Bases: alternat.generation.base.action_data_handler.ActionDataHandler

Rule for processing OCR data from driver.

Parameters: ActionDataHandler ([type]) – Base class for rule.

Initialize the handler with input data and confidence threshold (if available)

Parameters

input_data ([type]) – Data from driver.
confidence_threshold (float, optional) – Confidence threshold to filter OCR with low threshold, defaults to None (Driver config default)
ocr_filter_threshold (float, optional) – Confidence threshold to filter OCR data based on line height ratio to image height.

apply(interim_result: dict) → dict[source]

Process intermin result from previous rules in the chain and run OCR rule.

Parameters: interim_result (dict) – Intermediate results from previous rules in the chain.
Returns: [description]
Return type: dict

has_data() → bool[source]

Checks whethere OCR data is avalible in the input data.

Returns: [description]
Return type: bool

process_ocr() → dict[source]

Process the OCR data from the driver and filter it on the basis of line confidence threshold value and the ratio of line height to image height. Based on the configuration also invokes alternat clustering implementation (default to True)

Returns: [description]
Return type: dict

Caption Rule

class alternat.generation.rules.caption_handler.CaptionDataHandler(input_data: dict, confidence_threshold: float = None)[source]

Bases: alternat.generation.base.action_data_handler.ActionDataHandler

Rule for processing caption data from driver.

Parameters: ActionDataHandler ([type]) – Base class for rule.

Initialize the handler with input data and confidence threshold (if available)

Parameters

input_data (dict) – Data from driver.
confidence_threshold (float, optional) – Confidence threshold to filter captions with low threshold, defaults to None (Driver config default)

apply(interim_result: dict) → dict[source]

Process intermin result from previous rules in the chain and run caption rule.

Parameters: interim_result (dict) – Intermediate results from previous rules in the chain.
Returns: [description]
Return type: dict

has_data() → bool[source]

Checks whethere caption data is avalible in the input data.

Returns: [description]
Return type: bool

Label Rule

class alternat.generation.rules.label_handler.LabelDataHandler(input_data, confidence_threshold: float = None)[source]

Bases: alternat.generation.base.action_data_handler.ActionDataHandler

Rule for processing label data from driver.

Parameters: ActionDataHandler ([type]) – Base class for rule.

Initialize the handler with input data and confidence threshold (if available)

Parameters

input_data ([type]) – Data from driver.
confidence_threshold (float, optional) – Confidence threshold to filter labels with low threshold, defaults to None (Driver config default)

apply(interim_result: dict) → dict[source]

Process intermin result from previous rules in the chain and run label rule.

Parameters: interim_result (dict) – Intermediate results from previous rules in the chain.
Returns: [description]
Return type: dict

has_data() → bool[source]

Checks whethere label data is avalible in the input data.

Returns: [description]
Return type: bool

Library¶

class alternat.generation.generator.Drivers[source]

Bases: object

Driver name for alternat Library.

GOOGLE = 'google'

MICROSOFT = 'azure'

OPEN = 'opensource'

class alternat.generation.generator.Generator(driver_name: str = None)[source]

Bases: object

Generator class to implement alternat Library.

Raises

InvalidGeneratorDriver – Driver Invalid
InvalidGeneratorDriver – Driver Invalid
InvalidGeneratorDriver – Driver Invalid
OutputDirPathNotGiven – Output director path is not given.

Returns

[description]

Return type

[type]

Initializes generator with driver to use.

Parameters: driver_name (str, optional) – Name of the driver], defaults to None (opensource)
Raises: InvalidGeneratorDriver – Driver name is invalid or not implemented.

ALLOWED_DRIVERS = ['opensource', 'azure', 'google']

DEFAULT_DRIVER = 'opensource'

generate_alt_text_from_base64(base64_image: str)[source]

Generates alt-text from base64 image string.

Parameters: base64_image (str) – base64 image string
Returns: [description]
Return type: [type]

generate_alt_text_from_file(input_image_path: str, output_dir_path: str)[source]

Generates alt-text from file on disk.

Parameters

input_image_path (str) – Path to image to be processed.
output_dir_path (str) – Path to directory where the results needs to be saved.

Returns

[description]

Return type

[type]

get_config()[source]

Get the generator level config in the form of JSON.

Returns: [description]
Return type: [type]

get_current_driver()[source]

Get the current driver.

Returns: [description]
Return type: [type]

get_driver_config()[source]

Get the driver config in the form of JSON. Retreives public members [name: value] pair from the driver config class.

Returns: [description]
Return type: [type]

set_config(conf)[source]

Sets the generator level configuration parameters passed via JSON.

Parameters: conf ([type]) – Generator configuration parameters with values.

set_driver_config(conf: dict)[source]

Sets the driver config parameters using the JSON passed. There is one-to-one mapping between key in json and driver class public members.

Parameters: conf (dict) – Configuration JSON to set the driver configuration.

class alternat.collection.collector.Collector[source]

Bases: object

process(url: str, output_dir_path: str, download_recursive: bool = False, collect_using_apify: bool = False)[source]

Collects image from the url into the output directory

Parameters

url (str) – [description]
output_dir_path (str) – [description]
download_recursive (bool, optional) – [description], defaults to False
collect_using_apify (bool, optional) – [description], defaults to False