Tesseract: c-api return null PageIterator (Empty page) on rotated image

[reubano]

image: eurotext.tif

reubano@tokpro [~]⚡ convert eurotext.tif -rotate 3 +repage eurotext_03.tif
reubano@tokpro [~]⚡ convert eurotext.tif -rotate 5 +repage eurotext_05.tif

bug.py

#!/usr/bin/python
# -*- coding: utf-8 -*-

from __future__ import print_function, division

from os import path as p, environ
from ctypes import (
    CDLL, POINTER, Structure, c_char_p, c_bool, c_int, c_float, byref)

from ctypes.util import find_library

LIBTESS = find_library('libtesseract.dylib')
LIBLEPT = find_library('liblept.dylib')
TESSDATA_PREFIX = environ.get('TESSDATA_PREFIX', '/opt/local/share')


class TessBaseAPI(Structure):
    pass


class Pix(Structure):
    pass


class TessPageIterator(Structure):
    pass


def create_tess_api(prefix=TESSDATA_PREFIX, lang='eng'):
    tesseract = CDLL(LIBTESS)
    leptonica = CDLL(LIBLEPT)
    base_api = POINTER(TessBaseAPI)
    argtypes = [base_api, c_char_p, c_char_p]

    tesseract.TessBaseAPICreate.restype = base_api
    tesseract.TessBaseAPIInit3.argtypes = argtypes
    tesseract.TessBaseAPIInit3.restype = c_bool
    tesseract.TessBaseAPISetImage2.restype = None
    tesseract.TessBaseAPISetImage2.argtypes = [base_api, POINTER(Pix)]
    tesseract.TessBaseAPIAnalyseLayout.argtypes = [base_api]
    tesseract.TessBaseAPIAnalyseLayout.restype = POINTER(TessPageIterator)
    tesseract.TessPageIteratorOrientation.argtypes = [
        POINTER(TessPageIterator), POINTER(c_int), POINTER(c_int),
        POINTER(c_int), POINTER(c_float)]

    tesseract.TessPageIteratorOrientation.restype = None

    api = tesseract.TessBaseAPICreate()
    tesseract.TessBaseAPIInit3(api, prefix, lang)

    leptonica.pixRead.argtypes = [c_char_p]
    leptonica.pixRead.restype = POINTER(Pix)
    return tesseract, leptonica, api

def get_orientation(tesseract, leptonica, api, path, mode=1):
    tesseract.TessBaseAPISetPageSegMode(api, mode)
    pix = leptonica.pixRead(path)
    tesseract.TessBaseAPISetImage2(api, pix)
    it = tesseract.TessBaseAPIAnalyseLayout(api)

    if it:
        orientation, direction, line_order = c_int(), c_int(), c_int()
        skew = c_float()

        tesseract.TessPageIteratorOrientation(
            it, byref(orientation), byref(direction), byref(line_order),
            byref(skew))

        print('%s: %s' % (path, orientation.value))

if __name__ == '__main__':
    for path in ['eurotext.tif', 'eurotext_03.tif', 'eurotext_05.tif']:
        tesseract, leptonica, api = create_tess_api()
        orientation = get_orientation(tesseract, leptonica, api, path)

reubano@tokpro [~/Documents/Projects/tesseract]⚡ python bug.py 
eurotext.tif: 0
eurotext_03.tif: 0
Empty page!!

CR: https://github.com/sirfz/tesserocr/issues/5

[zdenop]

please follow FAQ#rules-and-advices

[reubano]

It would be helpful to have a CONTRIBUTING file with that information so that github will show it on the issues page.

[reubano]

Thanks to pyocr I figured it out!

# ... /snip
# https://github.com/jflesch/pyocr/blob/master/src/pyocr/libtesseract/tesseract_raw.py
class OSResults(Structure):
    _fields_ = [
        ('orientations', c_float * 4),
        ('scripts_na', c_float * 4 * (116 + 1 + 2 + 1)),
        ('unicharset', c_void_p),
        ('best_orientation_id', c_int),
        ('best_script_id', c_int),
        ('best_sconfidence', c_float),
        ('best_oconfidence', c_float),
        ('padding', c_char_p * 512),
    ]

# ... /snip

def create_tess_api(prefix=TESSDATA_PREFIX, lang='eng'):
    # ... /snip
    tesseract.TessBaseAPIDetectOS.argtypes = [base_api, POINTER(OSResults)]
    tesseract.TessBaseAPIDetectOS.restype = c_bool
    # ... /snip

def get_orientation(tesseract, leptonica, api, path, mode=0):
    tesseract.TessBaseAPISetPageSegMode(api, mode)
    pix = leptonica.pixRead(path)
    tesseract.TessBaseAPISetImage2(api, pix)
    osr = OSResults()
    it = tesseract.TessBaseAPIDetectOS(api, byref(osr))

    if it and osr:
        orientation, direction, line_order = c_int(), c_int(), c_int()
        skew = c_float()

        tesseract.TessPageIteratorOrientation(
            it, byref(orientation), byref(direction), byref(line_order),
            byref(skew))

        print('%s: %s' % (path, osr.best_orientation_id))
        print('confidence: %s' % osr.best_oconfidence)

[amitdo]

@reubano
I added 'tesserocr' to the AddOns wiki.
https://github.com/tesseract-ocr/tesseract/wiki/AddOns#tesseract-wrappers

[zdenop]

I added this example to APIExample

[amitdo]

It would be helpful to have a CONTRIBUTING file with that information so that github will show it on the issues page.

@reubano, thanks for the suggestion.
Now we have it! :-)