OCR API Credit Card Scanning Support

Hello! I am trying to parse a picture of a physical credit card. When I use the API, the JSON result is either completely blank or way off. Is this a known limitation? For example:

I would add that when I use the Online OCR API, and set Use OCR Engine 2 to true (which appears to not be production ready), it mostly works. Things like 6’s are sometimes recognized as b’s but otherwise it’s close.

Thanks!

Your image in the forum post is missing.

To clarify: So you mean it does not work with engine 1, but it works fine with engine 2?

Ah sorry the image didn’t paste before:
https://images.app.goo.gl/ojxZ3cx6JjF4GwZq6

(Save the image and upload instead of using the url)

It works better with engine 2, but not perfectly. When I set “do receipt scanning” to true it works even better. But with engine 1, with any parameters, it doesn’t pick up a single number from the credit card.

It works better with engine 2, but not perfectly.

Hmm… what does wrong with engine 2? In my test it works perfectly:

Sorry that was a bad example - here is picture of a physical credit card (dummy card):
https://images.app.goo.gl/rTdC3cP45AWDvzaN7

Since the image has already a good size, it is better to not use the auto-enlarge feature. If you uncheck this box, then you get:

This result is better. “6” is still confused with “b”, but this is something you can fix by a simple string replacement in the OCR result (replace all b with 6). What remains a problem is that “1171” is detected as “1111”.

We can do the string replacement for b and 6 - but we don’t really see a path forward for the 7 and 1.

Since engine 2 is still a major improvement, we would like to swap out engine 1 in our app. However it’s my understanding that PDFs are not supported by engine 2 - a feature that we need at a minimum. What do you recommend for someone that wants the improved ocr capabilities of 2, but still needs PDF scanning?

Thanks!

What do you recommend for someone that wants the improved ocr capabilities of 2, but still needs PDF scanning?

I recommend that, if possible, you just wait until September :wink: … we are working on exactly this feature currently.