Docling with windows + firewall

[pbonito]

Hi team,
I'm trying to install docling on a windows machine with limited connectivity to internet.
I sorted it out the error of model dependencies providing artifacts_path to pipeline. Now I'm struggling with OCR.
With EasyOCR pipeline fails with following error:

detector_path = self.getDetectorPath(detect_network)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\....\Lib\site-packages\easyocr\easyocr.py", line 253, in getDetectorPath
    download_and_unzip(self.detection_models[self.detect_network]['url'], self.detection_models[self.detect_network]['filename'], self.model_storage_directory, self.verbose)
  File "C:\....\Lib\site-packages\easyocr\utils.py", line 628, in download_and_unzip
    urlretrieve(url, zip_path, reporthook=reporthook)

Looks like trying to download some models, not sure how this can be sorted.

With tessercat I have an error while trying to install tesseract-ocr:

Using tesserocr-2.7.1.tar.gz
error: subprocess-exited-with-error

× Getting requirements to build wheel did not run successfully.
 │ exit code: 1
 ╰─> [58 lines of output]
     Failed to extract tesseract version number from: tesseract v5.4.0.20240606

      leptonica-1.84.1

       libgif 5.2.1 : libjpeg 8d (libjpeg-turbo 3.0.1) : libpng 1.6.43 : libtiff 4.6.0 : zlib 1.3 : libwebp 1.4.0 : libopenjp2 2.5.2

      Found AVX512BW

      Found AVX512F

      Found AVX512VNNI

      Found AVX2

      Found AVX

      Found FMA

      Found SSE4.1

      Found libarchive 3.7.4 zlib/1.3.1 liblzma/5.6.1 bz2lib/1.0.8 liblz4/1.9.4 libzstd/1.5.6
     Supporting tesseract v3.04.00

Looks it is looking for very old version.

Thanks in advance for your help.

[dolfim-ibm]

EasyOCR models

In this scenario, we suggest downloading the model files and point the code to its location.

This can be done with the argument pipeline_options.ocr_options.model_storage_directory = "PATH" see https://github.com/DS4SD/docling/blob/main/docling/datamodel/pipeline_options.py#L36.

Reading the EasyOCR docs, you should also be able to set the EASYOCR_MODULE_PATH ENV variable (we didn't test it). https://www.jaided.ai/easyocr/documentation/

Tesseracr via tesserocr

We also noticed that the version on pypi.org is not always linking to the correct Tesseract version. In the Docling docs we outline the possibility of installing the tesserocr binding library from sources, which will correctly link to the version installed on your system. https://ds4sd.github.io/docling/installation/

pip uninstall tesserocr
pip install --no-binary :all: tesserocr

[pbonito]

EasyOCR, models can be downloaded from here or here.

For tesseract even following your instruction I have the same error.

[saschacontes]

Please try to follow the tesseractocr installation for Windows:
https://github.com/sirfz/tesserocr and install the version that matches your Python from here
https://github.com/simonflueckiger/tesserocr-windows_build/releases