Downloads
v0.3.0 - move training process into Tesseract's new tesstrain.sh system.v0.2.2 - add training on more ligatured forms & glyphs, tweak dictionaries.v0.2.1 - add training on various punctuation marks and new fonts.v0.2.0 - fix use of Tesseract character blacklist, vastly improving accuracy.v0.1.0 - rebuild training under stable environment.v0.1.0-alpha2 - add training on bold and italic font variants.v0.1.0-alpha1 - initial training file prerelease.
Instructions: OS X / Linux / Windows
Code
- latinocr-lat - The final training process for lat.traineddata. Includes results from latinocr-lattraining process.
- latinocr-lattraining - Rules and tools to deterministically generate all prerequisites for the final training process.
- latinocr-lattestfodder - Latin page scans and ground truth text for testing OCR accuracy.
- tesseract_latinocr_docker - a Dockerfile for building lat.traineddata from scratch.
