You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Tesseract 4 generates the required property textangle with the old OCR engine, not when using LSTM. See ocropus/hocr-tools#148 for examples and more details.
The text was updated successfully, but these errors were encountered:
stweil
changed the title
Missing information on text orientation (hOCR property textrotation) with LSTM
Missing information on text orientation (hOCR property textangle) with LSTM
Mar 8, 2019
Interesting. This should already write the orientation angle (i.e. multiple of 90°) on current master. And it uses PageIterator.Orientation(), which is not related to (legacy) OSD, but the normal page segmentation included in PSM_AUTO, PSM_AUTO_OSD, PSM_AUTO_ONLY and PSM_SPARSE_TEXT_OSD. So it should work with any LSTM model loaded. I wonder why this issue even exists. (Have not tried it myself yet.)
But then, why does the code only use orientation value, but not add the skew angle?
@stweil Sure I'll do the same thing in #2815 as well.
Tesseract 4 generates the required property
textangle
with the old OCR engine, not when using LSTM. See ocropus/hocr-tools#148 for examples and more details.The text was updated successfully, but these errors were encountered: