how to make box file for tesseract

[Jingjing Lin]

I'm very confused about how to prepare text for further training tesseract. I don't think the page below gives any useful information about this:

https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract-4.00#making-box-files

Shouldn't there be a process where we input the correct text? Or is it that tesseract can just do everything for us? That will be kind of insane since tesseract makes mistake?

[Mox Betex]

You don't have to manually create .box files.

Use OCR-D for training https://github.com/OCR-D/ocrd-train

In data/ground-truth folder you put tif/gt.txt files and when you run make training it will generate box files.

For every tif image you write correct text in gt.txt file, nothing else.

Look at the https://github.com/OCR-D/ocrd-train/blob/master/ocrd-testset.zip for example of tif/gt.txt files.