The alpha2 release of OCRopus is now available from
www.ocropus.org
(follow the "Downloads" link).
In terms of codebase, this release largely achieves our goals for the
beta release, but we are calling it alpha2 because we want to do
additional training and tuning.
Here is a summary of changes:
* fixed dependency problems in jam
* implement OCR output in lua (in particular, hOCR output
routines)
* reduced memory usage of layout analysis
* xy-cut layout analysis
* Voronoi-based layout analysis (donated by K. Kise)
* Otsu binarization (in addition to our own fast Sauvola)
* script-based automatic EM training
* OCRopus and libraries callable directly from Python/NumPy
* replacement of ocrocmd by customizable Lua-scripts
* new beam search decoder
* graphical logging and debugging facility
* various bug fixes
There is also a growing amount of documentation on
www.ocropous.org
under "Documentation".
Cheers,
Thomas.