v0.20.1

Changed:

• ocrd-tool.json: remove configs/ as processor resource → no more directory resources
• Dockerfile: supplant configs/ resource for standalone CLI by pre-installing in tessdata

Added:

• ocrd-tool.json: add model resources for all Tesseract languages and scripts

v0.20.0

Changed:

• adapt to (and require) ocrd>=3.0 – allows running
  • with pages in parallel (OCRD_MAX_PARALLEL_PAGES) in tandem with METS Server
  • with page timeout (OCRD_PROCESSING_PAGE_TIMEOUT)
  • with page failure fallback copycat (OCRD_MISSING_OUTPUT=COPY), new default is SKIP instead of ABORT (now via --debug)
  • with page completion re-runs (OCRD_EXISTING_OUTPUT=SKIP), which is the new default instead of ABORT (now via --overwrite)
• switched to pyproject.toml build, tracking version via ocrd-tool.json

Added:

• more test coverage (esp. modes w/o METS Server, METS caching, instance-caching, page-parallel)
• Docker image includes preconfigured ocrd-all-tool.json for these processors

Fixed:

• no more logging side effects between tests

v0.19.1

Fixed:

• Correct version in ocrd-tool.json

v0.19.0

Fixed:

• segment*/recognize: delegate process instead of overwrite for docstring
• segment*/recognize: more robust polygon handling

Changed:

• more and more concrete tests
• 🔥 require Shapely v2
• Update tesseract to 5.4.1, #214

v0.18.0

Changed:

• tesseract and tesserocr included as submodules, installable via make instal-tesser{act,ocr}, #197
• Updated docker setup accordingly, #197

v0.17.0

Fixed:

• segment/recognize: fix shrink_polygons
• segment/recognize: fix reinit scope (for xpath_model and auto_model)
• CI: test multiple Python versions independent of ocrd/core image
• CI: speed up build for EOL Python 3.6
• CI: chmod o+w tessdata directory of PPA/OS Tesseract
• deps-ubuntu: allow installation of PPA Tesseract to fail (for newer OS)

Changed:

• adapted to Shapely v2
• *: inherit from recognize (but override logger)
• segment*: delegate constructor instead of wrapping instance
• requires ocrd==2.48

v0.16.0

Changed:

• require newer OCR-D/core to include OCR-D/core#934, #188
• no more need to set TESSDATA_PREFIX
• improved and up-to-date README

v0.15.0

Added:

• binarize: dpi numerical parameter to specify pixel density, #186
• binarize: tiseg boolean parameter to specify whether to call tessapi.AnalyseLayout for text-image separation, #186

Changed:

• regonize: improved polygon handling, #186
• resources: proper support for moduledir, companion to OCR-D/core#904, #187

v0.14.0

Changed:

• list all resources in the ocrd-tool.json, #184, OCR-D/core#800
• custom --list-resources handler, #176

v0.13.6

Fixed:

• segment/recognize: no find_tables when already looking for cells

Changed:

• segment/recognize: add param find_staves (for pageseg_apply_music_mask)
• segment/recognize: 🔥 set find_staves=false by default