OCR-D/ocrd_all v2025-04-11

on GitHub

Changed:

• Disable ocrd_detectron2 and ocrd_segment for OSX, #460

cor-asv-ann eb47a54..bc6a590

Release: v0.2.0

• CD PyPI: use latest packaging to fix metadata issue with twine
• fix license string to conform to PyPI req
• CD PyPI: use Pythonic base image
• docker: add vars
• Partial revert of "CI: install hunspell for ocrd-cor-asv-ann-mark test"…
• CI: differentiate store_test_results / store_artifacts
• requirements_test: fix typo (xdist instead of isolate)
• CI: reduce workload further, isolate tests to free memory
• CI: install hunspell for ocrd-cor-asv-ann-mark test
• tests: reduce workload of combinations to fit into 1h without GPU
• CI: cache model download, increase timeout
• CI: forgot to prefix model download directory
• CI: forgot to stage actual removal of Orb
• 📦 0.2.0
• CI: switch from Orb to CLI for Codecov
• CI: change codecov orb version
• setup.py → pyproject.toml (w/ dynamic setuptools-ocrd versioning)
• ocrd-tool.json: add 2 model files as examples
• add coverage, add tests to CI and pypi to CD
• docker: update+improve…
• increase default batch size for better GPU utilisation
• autosized generators: make work with Keras 2.3.x (stateful metrics)
• train: add model checkpointing callback
• seq2seq: be robust if EOL is missing in binary (prob/confmat) line input
• seq2seq: cover non-string line input (prob/confmat) in fallback
• transcode: simplify updating sequence, improve logging…
• transcode: pass all input TextEquiv alternatives…
• fix charmap kwarg init
• transcode test: abort on failure
• add test for transcode only, cover more fileGrps and parameters, check and compare confidences
• transcode: also remove lower hierarchy levels for consistency
• transcode: utilise .parent_object_, simplify
• tests: eval all against 'MULTI', test 'mark' separately, assert equal CER for both 'eval' strategies
• replace editdistance with rapidfuzz
• add tests
• migrate all processors to ocrd>=3.0
• CircleCI: separate CD and CI

docstruct 004e6ec..f8c4e46

• Merge pull request #2 from bertsky/migrate-ocrd-v3

format-converters fa8b4b5..0894801

• Merge pull request #25 from stweil/PEP8

ocrd_anybaseocr 0c67107..0aa1d57

Release: v2.0.0

• 📦 v2.0.0
• 📝 changelog
• Merge pull request #113 from bertsky/v3-api
• ocrd-tool: don't restrict resource_locations
• Be defensive about uninitializing possibly uninitialized model and label_mapping
• 3.8 typing
• self.reset at the end of process_workspace, del self.model in shutdown
• move models to root of package
• adapt tests
• rewrite with process_workspace and deferred METS modification
• rely on resolve_resource for model path
• use contextmanager to open pickled label mapping
• Merge branch 'v3-api' of https://github.com/OCR-D/ocrd_anybaseocr into v3-api
• remove redundant model-exists check
• CI: larger resource_class
• Merge branch 'v3-api' of https://github.com/OCR-D/ocrd_anybaseocr into v3-api
• pickle was ignored
• add model to package data
• reintroduce layout-analysis
• remove constants
• convert to src-layout
• remove old mets files
• remove pix2pixhd
• fix setuptools packages
• CI: remove references to models
• adapt cropper and test
• remove everything that is not cropping-related
• convert cropper to v3 API, setup.py -> pyproject.toml

ocrd_cis db65d7f..5cf22f5

Release: v0.2.0

• all: adapt to more Numpy and Shapely changes/deprecations
• get rid of remaining python-levenshtein (in favour of rapidfuzz)
• all: migrate importlib instead of EOL pkg_resources via ocrd_utils
• all: migrate to OCR-D v3 API (including METS Server and page multiprocessing)
• all: some code style, type hints and optimization
• tests: simplify, update, improve, test serially as well as page-parallel
• CI: ditch CircleCI, add Github Actions
• ocrd-tool.json: add dockerhub=ocrd/cis
• docker: add labels metadata conforming to spec
• docker: preinstall shortcut ocrd-all-tool.json via ocrd-tool.json
• setup.py → pyproject.toml, versioning based on ocrd-tool.json
• add continuous deployment for Dockerhub
• add continuous deployment for PyPI

ocrd_detectron2 218e0b5..3e5f838

Release: v0.2.0

• relax setuptools version requirement
• update setuptools to fix detectron2 build
• CI: remove Python 3.11 (does not work with detectron2)
• Update CHANGELOG.md
• Merge pull request #31 from bertsky/migrate-ocrd-v3

ocrd_froc 5a5d774..323a6ba

Release: v1.0.1

• 📦 v1.0.1
• 📝 changelog
• Merge pull request #20 from bertsky/patch-1
• 📦 v1.0.0
• 📝 changelog
• Dockerfile: COPY ocrd-tool.json to avoid symlink
• make docker: podman does not support --progress
• CI/GHA: actually run make test
• CI: docker
• Dockerfile: set locale/XDG variables, use COPY . . and .dockerignore
• Dockerfile: base on ocrd/core-cuda-torch
• Update requirements-test.txt
• tests: use subtests.test instead of plain for loop
• unit tests
• processor: handle ocr_method=none vs ocr_method=adaptive correctly
• OCR-D processor: restrict to 1 worker due to single CUDA context
• Update pyproject.toml
• Convert to OCR-D v3

ocrd_pagetopdf 00baf15..c6b414d

Release: v2.0.1

• 📦 v2.0.1
• 📝 changelog
• add changelog
• improve multipage structmap extraction (more robust, more cases)
• tests: debug logging only for processors
• deps-ubuntu: add apt-get update; add deps-test
• Merge pull request #29 from OCR-D/missing-import-re