I've just upgraded from Tesseract 3.00 to 3.01, as the multi-thread safe update is important to me.
First thing I noticed is that on my test image, 3.01 is less accurate. It makes many mistakes that 3.00 did not make (on the exact same image).
The result is the same whether I use the new 3.01 English trained data, or the older 3.00 trained data (as I was using before).
Anyone know what's up? Is there a new setting I should know about that's been disabled by default now or something?
