-
|
I edited the box files created by the trainer and I noticed that each character box has the size of the entire image as a box. I had never edited these files, but since reading the various tutorials I saw that one of the operations I should do is to verify that the single letter has the box that surrounds it, I was perplexed. I tried to do both English and Japanese training, and the result is identical. Is it correct that this is so? What am I doing wrong? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
I think this documentation page from the official Tesseract site should answer your question. The new LSTM training just uses whole lines, so the old per-character boxes are no longer necessary. In other words you are doing it right and it is expected. |
Beta Was this translation helpful? Give feedback.
-
|
You are very kind. I really thank you very much |
Beta Was this translation helpful? Give feedback.
I think this documentation page from the official Tesseract site should answer your question. The new LSTM training just uses whole lines, so the old per-character boxes are no longer necessary.
In other words you are doing it right and it is expected.