Tesseract OCR training gives 'APPLY_BOXES' errors

ocr, tesseract

Solution

Double check your .tif file character spacing: the first error could be caused by two characters being too close, and producing one single bounding box (see also this thread).

The second error is a consequence of the first one, I suppose.

Problem

I'm training Tesseract 3 for the digital-7 font, and I'm getting some errors in the output (from stderr): ``` APPLY_BOXES: boxfile line 35/I ((735,7860),(737,7890)): FAILURE! Couldn't find a matching blob ``` And ``` APPLY_BOXES: Unlabelled word at :Bounding box=(7523,-910)->(7549,-906) ``` What do they mean? The command line I'm using is: ``` tesseract eng.digital-7.exp0.tif eng.digital-7.exp0 nobatch box.train ```

Original source