Unsupported image object when using numpy.ndarray image

[baudneo]

My only guess is that somehow pytesseract is trying to import numpy from a diff env and the pkgutil.from_loader command sets numpy_installed to False? I have pytesseract installed into a venv.

Logs

  '10/14/23 11:34:46.0690' ZoMi:API[183741] DEBUG ocr:243 -> image for tesseract: type(letter) =<class 'numpy.ndarray'> -- isinstance(letter, np.ndarray) =True
  '10/14/23 11:34:46.0691' ZoMi:API[183741] ERROR ocr:251 -> PYTESSERACT OCR ERROR: Unsupported image object

Code

 if pytesseract is not None:
                ts = time.perf_counter()
                # first image should be the whole plate, the rest should be extracted letters
                # fixme: tesseract cli gets proper results on index 1+ but here it does not.
                psm = 7
                try:
                    for idx, letter in enumerate(letters):
                        if idx == 1:
                            psm = 10
                        logger.debug(f"image for tesseract: {type(letter) =} -- {isinstance(letter, np.ndarray) =}")
                        tess = pytesseract.image_to_string(
                            letter, lang="eng", config=f"--psm {psm}"
                        )
                        ret_[
                            f"Tesseract_{idx if idx > 0 else 'FULL IMAGE'}"
                        ] = f"PSM={psm} ---> {tess}"
                except Exception as exc:
                    logger.error(f"PYTESSERACT OCR ERROR: {exc}")
                else:
                    logger.debug(
                        f"perf:{self.LP} PyTesseract took {time.perf_counter() - ts:.5f} seconds"
                    )

Pytesseract Code

from pkgutil import find_loader

numpy_installed = find_loader('numpy') is not None
if numpy_installed:
    from numpy import ndarray

def prepare(image):
    if numpy_installed and isinstance(image, ndarray):
        image = Image.fromarray(image)

    if not isinstance(image, Image.Image):
        raise TypeError('Unsupported image object')

[stefan6419846]

Did you check that numpy is available in the same environment as pytesseract? What does pip show numpy pytesseract yield in the current environment? Which Python version are you using?

[baudneo]

Hi, I nuked the venv and did a reinstallation of a new venv and everything is working as expected. Sorry about that, not sure what the issue was, but all is well now.

Thanks!