You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As things stand, multiple hypotheses for (lines of) characters will only be available through multiple runs of OCR in OCR-D. What is missing is a clever way to serialize a single text version out of them. cor-asv-ann has the means available to provide this disambiguation step.
The text was updated successfully, but these errors were encountered:
This is definitely an objective. I have long-standing plans to integrate the ideas of http://www.ccs.neu.edu/home/dongrui/ocr.html. But this may need some time to harmonize, as the basic NMT implementations are quite different, and this would add a completely complementary training/data regime.
As things stand, multiple hypotheses for (lines of) characters will only be available through multiple runs of OCR in OCR-D. What is missing is a clever way to serialize a single text version out of them.
cor-asv-ann
has the means available to provide this disambiguation step.The text was updated successfully, but these errors were encountered: