How to set Confidence in tesseract

[bergheil]

Dear Tesser-Expert,
in the forum I found some information about confidence value for a
recognized word as integer between 1 to 255 (1 is the best).
Is it possibile to select that value running tesseract?
I building a training for Micr-CMC7 number and I would like the 100%
accurancy, so I prefer that tesseract reject a word instead recognize
with a low confidence.
Thanks in advice. Saluti a tutti

[Albert Law]

Hi,

I would look through each letter in your word to get the word's confidence. So something like this:


ETEXT_DESC* results = tess->Recognize_all_Words();
for (int iter = 0; iter < results->count; iter++) {
const EANYCODE_CHAR* ch = &results->text[iter];
if (255 > ch->confidence) {
// TODO: reject the letter and thus the whole word associated with it
}
}


-
Albert

[Remi Thomas]

From my experience, each word letter get the same confidence value =
the word confidence.

Rémi

[Albert Law]

Hi Remi,

Mmm, I thought that was a bug.

[bergheil]

Thanks for answer. I forgot to say that I do not use API but just
command line.
Regards

Debian 4.0
Tesseract 2.03

On Mar 17, 7:32 pm, "Albert Law" <a...@snowbound.com> wrote: