Is there any free/open source OCR available that can be trained for new symbols and can also output the coordinates of symbol found in the target image? I have read that tesseract OCR can be trained, but can it give me coordinates after OCR? any example? I need the code/steps to train a ocr using an image that contains one sybmol. There are around 20 symbols each in one image to be trained. and then use the trained OCR to detect those sybmols in the target image and if found, then give coordinates too.
Asked
Active
Viewed 4,205 times
1 Answers
4
You can train Tesseract to recognize new symbols. The hocr format contains the coordinates of the recognized words.
https://github.com/tesseract-ocr/tesseract/wiki/Training-Tesseract
http://vietocr.sourceforge.net/training.html
https://github.com/tesseract-ocr/tesseract/wiki/Command-Line-Usage#hocr-output

nguyenq
- 8,212
- 1
- 16
- 16
-
Do you have any more detail on this? The links are old/dead. – Tiago Apr 17 '17 at 10:21