recognize not seen sequence of character

65 views
Skip to first unread message

Essam Zaky

unread,
Jan 29, 2017, 6:50:59 AM1/29/17
to tesseract-ocr
Dears

I have question

is stlm can recognize not seen sequence of characters ?


here it's example for what i mean

if I have filed contain number consists of five digits 

when i train the LSTM should i generate all combinations of digits and feed it to LSTM trainer or just i can give few samples cover all digits

all combinations like the following
00000
00001
00010
...
...
...
99999


and here there are few samples contain all digits
01234
56789
are the two samples enough or i must generate the all combinations as explained in the previous step


Best Regards



ShreeDevi Kumar

unread,
Feb 2, 2017, 2:06:42 AM2/2/17
to tesser...@googlegroups.com, Ray Smith
+ @theraysmith

Two samples are certainly not enough for LSTM training. You have not specified whether you are training for a specific font for this particular numeric field of 5 digits. In case you are planning to use this traineddata ONLY for this numbers field, I think, you should train with all 100,000 combinations, since LSTM prefers large number of samples of the kind it is going to encounter.

cc:ing Ray, for confirmation.

ShreeDevi
____________________________________________________________
भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com

--
You received this message because you are subscribed to the Google Groups "tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tesseract-ocr+unsubscribe@googlegroups.com.
To post to this group, send email to tesser...@googlegroups.com.
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/f997d921-97f1-4896-9725-b71331def675%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply all
Reply to author
Forward
0 new messages