re: forms, I find it easier to just convert a pdf to an image or screenshot it programmatically, and process it with a few custom OCR APIs that identify key/val

re: forms, I find it easier to just convert a pdf to an image or screenshot it programmatically, and process it with a few custom OCR APIs that identify key/value form structures. Unfortunately, I would say, but it works pretty well.