Ask HN: How to do end-user batch handwriting recognition in 2018?

I have 9000 scanned handwritten book pages and wonder about what I can expect in terms of service and quality when running OCR on that. I tested Google cloud vision on a single page and it gives fairly good results (like "enough words detected for at least indexing the pages"), but I would rather like to use free or open source software like tesseract-ocr (which however is not specialized for handwriting OCR).

Who has experience with contemporary handwriting recognition? I cannot imagine there are no good datasets/codes available in our mobile-driven world. Are there open training datasets or similar which one can use to setup a digitalization toolchain?