Using Tesseract OCR for some automation tasks, I stumbled upon the following issue:
Creating OCR Image #1/4 with maximally 97 lines from line 1 to 97
ParamsModel::Incomplete line
ParamsModel::Incomplete line
ParamsModel::Incomplete line
ParamsModel::Incomplete line
ParamsModel::Incomplete line ConvNL
As it turns out, this is caused by using a wrong version of tesseract models / „Tessdata“. You can find the official „Tessdata“ files on GitHub. As I installed Tesseract on macOS via Homebrew, I ended up with version 3.05.01 of Tesseract. Therefore I had to use an older version of the trained models. As stated in the project’s README, you can find those via the 3.04.00 tag (README says branch, but that is wrong):
Get language data files for Tesseract 3.04 or 3.05 from the 3.04 tree.
E.g. for the german model, you would end up with the following download link:
https://github.com/tesseract-ocr/tessdata/raw/3.04.00/deu.traineddata