v2.0.0

Changed:

• 🔥 Port the processors (ocrd-anybaseocr-crop and ocrd-anybaseocr-layout-analysis) to OCR-D/core v3 API, #112, #113
• Run unit tests under various v3-specific conditions, #112
• Update Keras model metadata, #113

Deleted:

• 🔥 Dropped ocrd-anybaseocr-dewarp, #112
• 🔥 Dropped ocrd-anybaseocr-deskew, #112
• 🔥 Dropped ocrd-anybaseocr-binarize, #112
• 🔥 Dropped ocrd-anybaseocr-block-segmentation, #112
• 🔥 Dropped ocrd-anybaseocr-textline, #112
• 🔥 Dropped ocrd-anybaseocr-tiseg, #112

v1.10.0

Fixed:

• Docker setup, #102

v1.9.0

Fixed:

• block-segmentation: fix type of th parameter

Added:

• Resources now described in ocrd-tool.json, #100

v1.8.2

Fixed:

• dewarping: IDs must not contain /, #94
• uploaded trained models

v1.8.1

Fixed:

• Import pix2pixhd as package, not submodule, #93

v1.8.0

Changed:

• Use the newly created pix2pixhd module from PyPI, #91, NVIDIA/pix2pixHD#297

v1.7.0

Fixed:

• Makefile/tests: fix tests, update to resmgr cwd semantics, add dewarp
• layout-analysis: use correct pageId
• tiseg/layout-analysis: use TF SavedFormat instead of HDF5
• dewarp/layout-analysis: load during init (setup instead of process)
• dewarp: fix image input (in-memory instead of file-based)
• dewarp: fix image output (resizing with better quality)
• dewarp: fix/update pix2pixHD for CPU-only and newer PyTorch
• dewarp: rename parameters (now resize_{mode,width,height})
• dewarp: fix oplevel region, update to ocrd_mets changes
• update requirements
• improve README

v1.6.0

Removed:

• ocrd-anybaseocr-block-segmenation was broken and caused spurious CI errors

v1.5.0

Changed:

• Rewritten ocrd-anybaseocr-block-segmentation and ocrd-anybaseocr-tiseg, #79

v1.4.1

Fixed:

• Re-introduce requirement on ocrd-fork-ocropy