OCR-D/ocrd_all v2025-04-08

on GitHub

Removed

• sbb_binarization: Integrated into eynollah as ocrd-sbb-binarize
• ocrd_im6convert: Same functionality with improvements available as ocrd-preprocess-image in ocrd_wrap

core 71c0c19..80eb58c

Release: v3.3.0

• 📦 3.3.0
• 📝 changelog
• Revert "Merge remote-tracking branch 'bertsky/resmgr-type-checking'"
• 📦 v3.2.0
• 📝 changelog
• Merge remote-tracking branch 'bertsky/resmgr-type-checking'
• 📝 changelog
• update spec to v3.37.0
• 📝 changelog
• Merge remote-tracking branch 'bertsky/update-validator'
• 📝 changelog
• fix workdir in Dockerfile.cuda-torch
• fix deps-tf1 requirements DIRSTACK fix
• 📦 v3.1.2
• 📝 changelog
• deps-tf1: fix 7a2c4c7
• 📦 v3.1.1
• 📝 changelog
• test_processor timeouts: more tolerant
• test_resmgr: avoid name clash with preconfigured ocrd-dummy
• update ocrd-all-tool.json
• Dockerfile: repeat ARG BASE_IMAGE so it can be referenced below
• nw integration test: pass build ctxt for ocrd_core_test everywhere
• nw integration test: never pull ocrd_core_test (wait for build in PS service)
• deps-conda: new target for non-ubuntu platforms
• deps-cuda: expose get-conda independently, re-use if present, otherwise use Micromamba for current OS/platform
• docker*: pass build args, also tag with version if gitref is release
• Dockerfile*: always provide defaults for ARG, use upper case AS
• docker-{torch,tf{1,2}}: smoke-test build, too
• deps-{torch,tf{1,2}}: ensure core's requirements.txt is compatible
• 📦 v3.1.0
• 📝 changelog
• Merge branch 'update-spec'
• 📝 changelog
• ocrd-tool: add command 'dump-tools'
• adapt processor tests some more
• adapt processor tests
• Processor.download: apply config.OCRD_DOWNLOAD_INPUT no sooner than available, remove useless download_files kwarg
• 📦 v3.0.4
• 📝 changelog
• update assets and adapt tests to match
• tests.processor: increase timeout tolerance slightly
• run_processor: handle frozendict or empty parameter in profile log
• tests.processor: check profile log
• tests.processor: add case adding document-global file (in METS Server mode, too)
• tests.processor: add case adding document-global file
• tests: OcrdPage roundtrip w/ full assertions
• OcrdPage: re-generate
• fix c47ae77c (pass on namespaceprefix_)
• tests: add OcrdPage roundtrip
• mets server: fix init of 'page_id' in POST /file
• mets server: downgrade log messages INFO→DEBUG
• test_workspace: use different asset w/ loctype URL
• 📦 v3.0.3
• update assets and spec
• 📝 changelog
• Merge remote-tracking branch 'bertsky/fix-processpool-loglistener-shutdown'
• 📦 v3.0.2
• 📝 changelog
• fixup 39f76258: get queue type from mp.get_ctxt
• Processor.ProcessPoolExecutor: use queue.Queue instead of mp.Queue for QueueHandler
• Workspace.image_from_*: remove PIL workaround
• ocrd_network.cli.client processing: fix CLI arg order
• ocrd.decorators: update/improve docstring
• config.OCRD_MAX_PARALLEL_PAGES: improve docstring
• 📦 v3.0.1
• 📝 changelog
• Processor.ProcessPoolExecutor: start QueueListener only after executor to avoid thread interference
• Processor.DummyExecutor.shutdown: reset global processor instance, too
• CI: Use token authentication for pypi
• 📦 🔥 v3.0.0
• remove 3.8 breaking typing hints
• 📝 changelog
• Merge remote-tracking branch 'bertsky/new-processor-api-xpath-filter' into v3-release
• 📝 changelog
• Merge remote-tracking branch 'origin/initLogging-before-imports' into v3-release
• 📝 changelog
• Merge branch 'fix-help-output' into v3-release
• 📝 changelog
• Merge remote-tracking branch 'origin/master' into v3-release
• merge master
• 📦 v3.0.0b7
• Processor parallel pages: log via QueueHandler in subprocess, QueueListener in main
• Processor: replace weakref with del to trigger shutdown
• initLogging: call disableLogging if already initialized and force_reinit
• Processor: update max_workers docstring
• logging: increase default root (not ocrd) level from INFO to WARNING
• setOverrideLogLevel: override all currently active loggers' level
• disableLogging: remove all handlers, reset all levels
• initLogging: only add root handler instead of multiple redundant handlers with propagate=false
• initLogging: do not remove any previous handlers/levels
• tests: prevent side effects from ocrd_logging
• 📦 v3.0.0b6
• Merge pull request #23 from bertsky/new-processor-api-process-worker
• Processor.process_page_file: avoid process_page_pcgts() if OCRD_EXISTING_OUTPUT!=OVERWRITE
• ocrd.cli.workspace: use physical_pages if possible, fix default output_field
• OcrdMets.get_physical_pages: cover return_divs w/o for_fileIds for_pageIds
• ocrd.cli.workspace: assert non-server in cmds mutating METS
• 📝 improve b5 changelog
• 📦 v3.0.0b5
• update OcrdPage from generateds
• OcrdPage: add PageType.get_ReadingOrderGroups()
• tests: make sure ocrd_utils.config gets reset whenever changing it globally
• test processors: no need for 'force' kwarg anymore
• Processor / Workspace.add_file: always force if config.OCRD_EXISTING_OUTPUT==OVERWRITE
• lib.bash input-files: do not try to validate tasks here (impossible to get right with required parameters, and now covered by wrapped Processor.verify() already)
• lib.bash input-files: pass on --mets-server-url, --overwrite, and parameters (necessary for required params)
• lib.bash: fix errexit
• run_processor: be robust if ocrd_tool is missing steps
• Processor.verify: check output fileGrps as well (or OCRD_EXISTING_OUTPUT=OVERWRITE|SKIP or disjoint --page-id)
• Processor.process_workspace(): do not show NotImplementedError context if fallback process() raises anything itself
• ocrd.cli.validate tasks: pass on --mets-server-url, too
• ocrd.cli.bashlib input-files: pass on --mets-server-url, too
• ocrd.cli.workspace server: add 'reload' and 'save'
• ocrd.cli.workspace: consistently pass on --mets-server-url and --backup (also, simplify)
• METS Server: also export+delegate physical_pages
• processor CLI: delegate --resolve-resource, too
• ocrd.cli.ocrd-tool resolve-resource: fix (forgot to print result)
• PcGts.Page.id / make_xml_id: replace '/' with '_'
• Processor.process_page_file / OcrdPageResultImage: allow None instead of AlternativeImageType
• Processor.verify: revert 5819c816 (we still have no defaults in json loaded from v2)
• add typing, extend docs
• test_processor: add test for force (overwrite) w/ METS Server
• test_mets_server: add test for force (overwrite)
• OcrdMetsServer.add_file: pass on 'force' kwarg, too
• 📦 v3.0.0b4
• 📝 changelog
• Workspace.reload_mets: fix for METS server case
• Processor: log when max_workers / max_page_seconds are in effect
• Processor.metadata_location: find location package prefix (necessary for namespace packages)
• 📦 v3.0.0b3
• ocrd.cli doc: don't rewrap description lists
• Processor: improve processing log messages
• add test for OcrdEnvConfig.reset_defaults()
• ocrd_utils.config: add reset_defaults()
• 📦 v3.0.0b2
• 📝 update readmes (esp. new config variables)
• 📝 update changelog
• test_processor: add legacy (v2-style) dummy case
• bashlib: re-add --log-filename, implement as stderr redirect
• Processor.generate_processor_help: forgot to include --log-filename
• Processor: fix 7966057f (deprecated passing of ocrd_tool or version via init)
• test-logging: also remove ocrd.log from tempdir
• disableLogging: re-instate root logger, to
• ClientSideOcrdMets: partial revert of 9b80ae17ef
• tess: adapt to wording of exceptions
• test processor timeout/parallel: avoid side effects to dummy tool json
• test_mets_server: use tmpdir to avoid side effects between suites
• Processor: fix ignore (negative/zero) cases for max_workers / max_page_seconds
• ClientSideOcrdMets: use same logger name prefix as server
• 📝 update changelog
• add tests for processor per-page timeout and parallelism
• Processor: add per-page timeouts and parallelism…
• Merge pull request #19 from OCR-D/new-processor-api-fix-editable
• adapt tests for OCRD_MAX_MISSING_OUTPUTS
• Processor.process_workspace: abort anyway if too many failures (OCRD_MAX_MISSING_OUTPUTS)
• generate_processor_help: avoid repeating docstrings from superclass
• 📦 v3.0.0b1
• Merge branch 'master' into new-processor-api
• Processor.process_page_file / OcrdPageResultImage: allow PageType instead of AlternativeImageType
• test_bashlib: allow testing prereleases successfully
• 📝 update changelog
• Merge pull request #14 from bertsky/new-processor-api-parameter-setup
• remove make *-workaround, we will not do that for v3+
• 📦 v3.0.0a2
• 📝 changelog
• processor CLI --debug: set all to ABORT (not just MISSING_OUTPUT)
• predefine union types OcrdFileType and OcrdPageType
• JsonValidator: fix deprecation warning (by actually checking instance)
• tests report.is_valid: improve output on failure
• dummy_processor: fix typos from logging
• OcrdPage: fix typeing typo
• Merge pull request #13 from OCR-D/validate-ocrd-tool-runtime
• Merge pull request #11 from OCR-D/ocrd-page-with-etree
• Merge pull request #10 from OCR-D/processor-logger
• Merge pull request #12 from bertsky/new-processor-api-input-file-errors
• Merge branch 'master' into new-processor-api
• s,PcGtsType,OcrdPage,
• fix 9cfd70cffcc (revert to wrong import order to avoid circle)
• fix 9cfd70cffcc
• ocrd_utils: forgot to export scale_coordinates at toplvl
• fix type assertion
• fix imports
• ocrd.cli: more fix module import order, export help cmd
• ocrd: more convenience imports
• Update CHANGELOG.md
• 📝 v3.0.0a1
• Merge branch 'processor-result-object' into new-processor-api
• update spec (with new ocrd_tool.schema)
• bashlib: fix ocrd__minversion test syntax
• fix ocrd_tool.schema.yml cardinality oneOf syntax, update spec
• Merge pull request #7 from OCR-D/bashlib-version-yak-shaving
• test_bashlib: use version verbatim
• Merge branch 'master' into new-processor-api
• CircleCI: use version 2.1
• Set VERSION to upcoming 3.0.0a1
• improve output in case of assertion failures
• hide/test expected deprecation warnings
• use up-to-date kwargs (avoiding old deprecations)
• adapt to ocrd-tool.json cardinality changes
• test_processor: adapt to Processor init changes
• Processor / ocrd-tool.json: 🔥 fileGrp cardinality checks…
• Processor/CLI decorator: 🔥 separate kwargs and constructor…
• mets_server: ClientSideOcrdMets needs OcrdMets-like kwargs (without deprecation)
• test_mets_server: fix arg vs kwarg
• docs:regenerated rST
• docs: add ocrd_network module
• docs: do/not exclude tests/src
• docs: fix relative VERSION path
• Processor init: deprecate passing workspace
• run_processor: fix typo
• process_page_pcgts: add (variadic) type checks
• Processor.process_workspace: add fileGrp assertions
• run_processor w/ mem_usage: pass as args tuple
• use more specific exception if parameters are invalid
• ocrd-tool CLI: pass class in context
• get_processor: also run setup if instance_caching
• generate_processor_help: include process_workspace docstring, too
• tests: adapt to new Processor init (override metadata/version/executable name)
• adapt to new Processor init (override metadata/version/executable name)
• autoload ocrd-tool.json and version from dist, executable name from entry point in stack
• test_workspace: adapt to image_from_* DPI and add assertions
• Workspace.image_from_*: annotate 'DPI' in result dict and ensure it's used in meta-data of resulting image
• Workspace.save_image_file: save DPI metadata, too
• Processor.process_page_pcgts: add kwargs and allow returning derived images
• Workspace.save_image_file: add kwarg file_path for predetermined local_filename
• test processor: adapt to self.workspace only during run_processor
• test processor: adapt to input_file_grp required
• test builtin ocrd-dummy: adapt to consistent filename
• test DummyProcessor: override process_workspace() by delegating to process() directly
• test DummyProcessor: adapt to new download default by setting download_files=False in tests (because they are not actually in the filesystem)
• run_processor: adapt to process→process_workspace
• DummyProcessor: re-implement via new process_page_*
• Processor: refactor processing API…
• ocrd_cli_wrap_processor: fix workspace arg (not a kwarg)
• Processor: add setup(), run once in get_processor()
• Processor: factor-out show_resource(), delegate to resolve_resource()
• workspace.download: fix typo in exception
• fix #274: no default -I / -O
• deprecate Processor.process()

eynollah bc9dddd..8080bd8

Release: v0.4.0

• 📦 v0.4.0
• 📝 changelog
• test_run: ensure exceptions are shown
• 'from PIL.Image import Image' causes an error when using Image.new(), and since Image is already imported, this line can be safely commented out.
• CI: use clash-free artifact name for report upload
• CI: fix coverage report calls
• CI: make coverage cfg work with both editable and dist install
• CI: forgot to (re-)enable verbose logging
• CI: (try to) store/upload coverage results
• binarization CLI: fix option checks, simplify to asserts, fix dir_in mode
• pytest: add tests for directory mode (layout+bin)
• pytest: add asserts for results, add binarization
• pytest: use subtests for various layout options, add coverage
• CLI: simplify option checks to asserts (also avoid stack trace)
• logger: fix type hint
• Merge remote-tracking branch 'origin/main' into v3-api-release-foreal
• Merge pull request #130 from qurator-spk/v3-api
• This commit enables the export of cropped text line images along with their corresponding texts from a Page-XML file. These exported text line images and texts can be utilized for training a text line-based OCR model.
• For the CNN-RNN OCR model, long text lines are split into two segments
• Resolved an issue in the OCR-D framework where dir_out received a None value
• Resolving test failure
• Merge remote-tracking branch 'bertsky/machine_based_reading_order_integration_fixes' into machine_based_reading_order_integration
• changing cnn ocr model name
• Provide OCR as an option to process a directory of XML files, incorporating layout and text line coordinates.
• adding scatter_nd inference
• fixed: machine based reading order cause tuple index out of range error if number of textregion is one.
• filter textregions without textline
• gha: try to free disk space
• Update Makefile
• new table detection model is integrated
• fixing IndexError by reading order detection
• updating requirements
• tests are passed - new models by the way should be uploaded
• textlines of drop capitals are connected to corresponding textline if possible otherwise they are inserted in corresponding textregion
• drop capital and marginals extraction is updated
• Update requirements.txt
• sbb_binarization is integrated into eynollah works in framework of ocrd - sbb_binarization in ocrd works for individual images by the way as standalone flowing from directory can be used now. For eynollah in ocrd framework I have added -light version as default parameter.
• resolve merge conflict of main branch with machine based reading order branch
• Resolve merge conflict of main and machine based reading order branch
• updating early layout inference for light version
• updating inference for early layout in the case of documents with number of columns bigger than 2
• updating
• textline segmentation is masked with drop capitals
• Merge branch 'machine_based_reading_order_integration' of https://github.com/qurator-spk/eynollah into machine_based_reading_order_integration
• fixing a bug occuring with reading order + Slro option with no patch textline model and thresholding artificial class
• binarization as a standalone command
• doing more multiprocessing in order to make the process faster
• textlines of textregions are extracted in a faster way + early layout for all documents is done with no patches model and on rgb input
• updating light version
• -light version need -tll to be enabled otherwise the process will be ended.
• updating light version features
• updating light version. Remove textlines or textregion contours inside a bigger one
• updating light version
• dilation of textregions and marginals are accomplished
• updating textregions dilation
• parametriyation in the case of textline contours dilation is accomplished
• updating textline dilation parameters
• updating dilation of textlines and text regions
• dilation of text regions without opencv
• updating light version
• update of light versions
• postprocessing of textline contour dilation + skip layout and reading order passed as an argument
• the most effective version of contours dilation without opencv and all at once
• updating scaling contours
• scaling contours without dilation
• passing number of columns as an argument
• avoiding double binarization
• adding option for textline detection in printspace
• resolving issue #110 in a better way
• writing drop capitals in xml output + and may resolve issue #110
• inference batch size debugged
• ignoring dpi check by light version
• making light version faster for 1 and 2 columns images
• inference with batch size bigger than 1
• update Makefile model location
• update pyproject.toml
• dtype = object in the case of length 1 arise error
• pyproject.toml may work for ocrd
• resolving #106 for pyproject.toml test
• testing pyproject.toml
• 1&2 column images only printspace
• updating 1&2 columns images + full layout
• new full layout model and early layout for 1&2 column images are integrated - light version
• ocr engine first integration
• machine based reading order as an argument
• machine based reading order & works for not full layout case
• machine based reading order inference & optimized algorithm
• machine based reading order inference with a variable batch size
• integration of machine based reading order detection

nmalign 1426dbc..1698cd8

Release: v0.0.5

• Merge pull request #1 from bertsky/migrate-v3

ocrd_anybaseocr 483848f..0c67107

Release: v1.10.0

• Merge remote-tracking branch 'origin/fix-ci-versions'

ocrd_calamari d7c30b5..416420f

Release: v2.0.1

• 📦 v2.0.1
• 📝 changelog
• fix glitch in Makefile from merge
• 📦 v2.0.0
• 📝 changelog
• Merge branch 'master' into calamari2
• docker: fix/update base image
• improve dockerfile
• simplify dockerfile
• add docker and build rules, improve dockerfile
• fix setuptools discovery
• workaround: run eagerly to prevent rare TF function stalls
• precompute predict_function
• require ocrd 3.0.3 (clean QueueListener shutdown)
• Merge remote-tracking branch 'bertsky/calamari2' into calamari2
• CalamariPredictor.shutdown: not from forked processors
• CI: fix indentation
• CI: store coverage results
• coverage: try to include subprocess results
• require calamari-ocr 2.3.1 (s11n of results via queue)
• install dev before normal requirements to avoid typing-extensions conflict
• require ocrd 3.0.0b7 (sub logging)
• test: no need for --isolate anymore
• test: default to supported model
• test: no need for fixture as dict
• test: proper shutdown w/ no side effects
• fix word/glyph level
• replace atexit with weakref, also add destructor
• Merge branch 'calamari2-subprocess' into calamari2
• disallow TF 2.12.0
• ocrd-tool.json: remove model default
• Makefile: export MODEL variable
• CI: wait longer for test without output
• ocrd-tool.json: fix 2.2 model URLs (zip→tar.gz)
• CI: test on smaller model
• Merge branch 'master' into calamari2
• update models (2.1→2.2 for Py3.11 re syntax)
• update requirements (Calamari 2.3, remove implied TF)
• update models (5→6 HDF5→SavedModel)
• use C2 deep3_fraktur19 model for testing
• add C2 models
• 📦 2.0.0
• expose device parameter, default to GPU
• test with pytest-isolate so TF memory can be freed between tests
• test configs: test metsserver independent of pageparallel
• voter cannot be parallel; shmem is faster
• convert images to input_channels (required TensorShape)
• wait for executor to fully initialize (so we get input_channels)
• 🔥 adapt to (and require) Calamari 2…
• CI: increase RAM
• no more need for model fixup
• let GPU memory grow by demand (instead of exclusive reservation)
• run prediction in bg thread (shared across pages to interleave CPU/GPU)
• aggregate all lines instead of per region to better utilise batched predictor
• require ocrd 3.0 and calamari-ocr 1.0.7
• tests: adapt to v3, overhaul and add caching+threading modes
• make test: no assumption on OCRD resource location
• adapt to ocrd>=3.0

ocrd_kraken 17cefc6..4309d91

Release: v1.0.1

• Merge pull request #46 from bertsky/docker-res
• 📦 v1.0.1
• 🔥 📦 v1.0.0
• Merge branch 'port-to-v3'

ocrd_neat 1982216..b9cb043

Release: v0.0.1

• add drop-columns options to tsv2tsv

ocrd_olena a02e937..78518b8

Release: v2.0.0

• Merge pull request #99 from bertsky/v3-python

ocrd_pagetopdf 7e8c860..00baf15

Release: v2.0.0

• PyPI CD: fix workflow
• CD GHCR: downcase GH group
• 📦 2.0.0
• 📦 0.2.0
• Merge pull request #28 from bertsky/master

ocrd_page2alto 8877e8f..da309e6

Release: v1.5.0

• :package v1.5.0
• 📝 changelog
• ocrd-tool.json: Set steps to postprocessing/format-conversion OCR-D/spec#261
• downgrade reqs to make py37 and py38 work
• iterate over regions non-recursively, skip_empty_line bug, invert check_border default
• don't check border by default

ocrd_segment 064b7a8..a5d3d62

Release: v0.1.24

• 📦 v0.2.0
• Merge pull request #69 from OCR-D/port-to-v3

ocrd_tesserocr dcbd522..c65b345

Release: v0.20.1

• 📦 v0.20.1
• docker: update to ocrd==3.3.0
• Dockerfile: supplant configs/ resource by pre-installing
• ocrd-tool.json: add all langs
• ocrd-tool.json: remove 'configs/' resource, add all scripts
• Dockerfile: no more /models alias
• CI: increase no-output timeout
• Merge pull request #216 from bertsky/migrate-ocrd-v3

ocrd_wrap fd4a2bc..83ab5f0

Release: v0.2.0

• Merge pull request #12 from bertsky/docker-res
• Update CHANGELOG.md
• 📦 0.2.1
• Update pypi.yml
• 📦 0.2.0
• Merge pull request #11 from bertsky/migrate-v3

workflow-configuration 63e9969..a5bffee

Release: 0.1.3

• Merge pull request #34 from bertsky/migrate-v3-pythonize