5 ms·

Can you elaborate? What is the use case?

• staticautomatic 9y ago
  Extraction from semi-structured documents. Reqs: - No absolute element positions - Able to handle a reasonable amount of skew - Able to handle photographs of documents (e.g. mobile) - Excellent support for repeating groups/elements - Around 95% accuracy without human verification
  • ocrcustomserver 9y ago
    Interesting, I have implemented some version of the above in the past (albeit not in an opensource/free solution). If you have a specific document/data extraction need in mind, you can email me a sample and I will let you know if I can help.
  • loxias 9y ago
    Similar to the other comment you received, I've also been working on some of the math necessary to do such a a thing! I have no idea how it would be productized there's so many things out there that do "good enough" jobs... One of my ideas was to extract structure from motion and use that to deskew and align.. (Can I also have a datadump of candidate files for testing?