v0.1.8

• extract tables as images (#517, #519)
• various dependency updates

v0.1.7

• added support for remote training data generation (#514)
• generate delft data (#510)
• various dependency updates

v0.1.6

• output training data files to a directory structure (#502)
• generate training data references for the name model, for the reference author names (#501)
• generate training data for the name model, for the author names in the header section (#493)
• generate training data for the table model (#492)
• refactored generate training data (#491)
• generate training data for the figure model (#489)
• generate training data for the citation model (#488)
• added missing fulltext training data ref section (#486)
• generate reference-segmenter training data (#485)
• generate training data for the fulltext model (#483)
• improve generate data tests (#482)
• generate affiliation-address training data (#481)
• generate header training data (#479)
• fixed segmentation acknowledgement training data gen (#477)
• generate segmentation training data using model (#476)
• added segmentation model training data generator (#472)
• added link to usage examples (#471)
• various dependency updates

v0.1.5

• fixed asset zip failing with assert on relative path (#469)

v0.1.4

• added Jupyter notebook support (#467)

v0.1.3

• fixed missing xslt and default config in pypi package (#464)
• fixed missing setup resources in python package (#463)

v0.1.2

• refactored python api (#459)
• improved python library documentation (#458)

v0.1.1

• push pypi package (#457)
• download to the user's home directory by default (#455)
• moved config to sciencebeam_parser/resources/default_config (#454)
• switched to sciencebeam-trainer-delft pypi release (#453)
• optionally replace text by cv graphic (#452)
• fix multiple affiliation markers (#451)
• various dependency updates

v0.1.0

• only preload cv and ocr models if enabled
• guess media type if empty
• (renamed sciencebeam to sciencebeam-pipelines; redirect to sciencebeam-parser)

v0.0.2

• preload models
• add word* support
• added /processReferences endpoint
• added /processHeaderDocument endpoint
• added /convert API endpoint
• added jats support
• fix graphic coords assert and division by zero
• match svg pages with uncommen page dimensions
• added ocr config params; configured sparse text ocr psm
• fixed ocr no image error with large image
• prevent concurrent tesser_api calls
• fixed local file path assert
• figure label ocr
• added libgl1 to docker image
• added poppler to docker image
• integrate layout parser for cv figure detection
• added download_if_url_from_alternatives
• fixed unnessary download of gzipped wapiti model
• avoid matching graphic svg
• use send file for zip with assets to avoid memory issues
• implement asset document
• implemented figure bounding box baseline
• refactored tei document
• split aff on other label
• added support for config env bool value
• made use_first_token_of_block configurable
• move formula out of paragraph
• improve segmentation fulltext consistency
• fixed app features context not passed to some models
• use underscore for model config keys
• implemented first and last name lookup feature
• add author aff markers as delimters
• switched default name header model to 0.6.0
• improve name header
• various dependency updates