Practical approach for minimal change is reading dotted numbers

18 views
Skip to first unread message

Bernard Trigaux

unread,
Mar 17, 2025, 11:02:22 AM3/17/25
to tesseract-ocr
Hi,

I have an appliance regulating chlore level in a pool.
Very basic need to read 2 numbers like on the photo here below :
regu.jpg
A first experimentation shows that "733" has been properly discovered and returned.
The "0%" however returned as "8%"
Tesseract-Run1.jpg

So in summary :
- no question of language, only digits and '%' to be recognized.
  (-c tessedit_char_whitelist=0123456789%)
- is it correct that I only need to retrain for those characters 0123... ?
- how to retrain : take as many photos as necessary of the appliance screen until I have all digits on a photo ? Then follow the process to prepare an increment of test data.

Is this approach ok ? Welcomed any hint !

Thans you !

Reply all
Reply to author
Forward
0 new messages