#bbuzz 2018: Mark Keinhörster – Scalable OCR pipelines using Python, Tensorflow and Tesseract