4

I would have to say this online OCR software that I was forced into and expected to build for medical documents. The problem was the scanned documents were so unreadable, crooked, and in dot matrix style, so there was really no way to do this.

Comments
  • 0
    Don't know when that was, but now Microsoft have a something called Form recognizer that's in preview, that looks promising and is made for reading scanned documents 🙂
  • 0
    Google Cloud Vision API works pretty well too
    Deep learning can sometimes still work when traditional machine learning would fail
  • 0
    True, well these scans you could barely read, that's the issue. And then it would need to parse the data in a certain format, like CSV and headers.
  • 0
    @zcoder form recognizer combined with logic apps from Azure can do that 🙂, I tried trashing it a few days ago, and it got it all except special handwriting, but that was without any training, so that was expected
Add Comment