TesseractOCR custom training

Question

I don't get it how am I supposed to tell Tesseract that this is actually a b c letters?

Appending makebox in cli, produced this file:

C 78 127 104 166 0
l 96 127 118 166 0
. 116 127 126 140 0
- 124 127 130 145 0
j 145 127 162 188 0
o 154 127 183 158 0
b 203 131 241 164 0

Answer 1

You can use the following link to know more about how to train tesseract. To correct the box files you can use jTessBoxEditor. Do keep in mind at times you might need to preprocess the images.