I've spent a while now refining various things with the Ancient Greek training.
The result is significantly better, not least due to the text2image tool,
specifically its' --exposure setting. It's compressed with xz, which reduced it
from 8.4MiB to 2.2MiB (xz is amazing).
Also attached is the complete build recipe (as a self-contained makefile) and
source files (grc-src.tar.xz). I reckon the files should live in
training/langdata/grc/, though I know Ray has some plans for how the training
data should be organised in the future. This is how I imagine things being
organised, anyway. Some of the files in grc-src.tar.xz are themselves
generated, using the tools in the git repo at
http://ancientgreekocr.org/grctraining.git, but the grc-src.tar.xz files are
appropriately modifiable and self-contained that I think it makes sense to host
them with the other training data.
Original issue reported on code.google.com by nick.wh...@durham.ac.uk on 26 Apr 2014 at 9:44
Original issue reported on code.google.com by
nick.wh...@durham.ac.uk
on 26 Apr 2014 at 9:44Attachments: