Yeah, I was thinking something along the lines of lots n lots of easy shitty data (ex. anyone who can take a picture), some pretty good data (ex. hand labeled receipts), some 100% reliable data (scraped/api) then some sort of system to correlate the 3, especially when prices match identically between receipt and api a fair sized database could create itself.

Also would need some sort of processing center to handle the many image processing requests, but maybe that could be done client side