API demo files give incorrect results

Hello there. I am experimenting with your service for inclusion in a couple products I am prototyping at the moment.

However, I notice strange results from your service, even on the demo file you mention “http://i.imgur.com/fwxooMv.png” -> in particular, - with version 1, the words “Gave Away” contain a non-ascii “a” character, and - with version 1 & 2, the word “AMAZON” is detected as “AMAMON” ?? Is this a known issue?

To resolve these and any similar issues, is it recommended that I do any particular image pre-processing (such as rotation, perspective adjustment etc, as per this article https://www.pyimagesearch.com/2014/09/01/build-kick-ass-mobile-document-scanner-just-5-minutes/ ) or do you take care of that in your api and so anything I do as preprocessing would be wasted? Thanks!

Preprocessing is usually not required for the ocr api with one exception:

Make sure that the text fonts are neither too small nor too large. The more “normal” the size, the better the OCR result. In your case you have an image with very large letters so uncheck to the auto-enlarge box (scale parameter). Then the result is 100% correct with engine2:

1 Like