Provide information about normalization
Opened this issue · 1 comments
mweidling commented
CER also depends on the model and OCR engines used (character set, punctuation, hyphenation char, newline, kind of quotation marks) -> some normalization between GT and Model output often needed
paulpestov commented
Currently we are displaying only the model name. We could enhance this with some additional info and move it somewhere more prominently.