Google cloud Vision API gives different results to drag and drop example

1
votes

I'm using the google vision (OCR) API (feature "TEXT_DETECTION") and I've noticed differences in the results.

The results given using Google's drag and drop page
differ to the results returned from an android app which calls the same API. The android example code is provided by Google on github (I have modified it to use text detection rather than labels API)

Using the same image, drag and drop returns perfect results but the android app which calls the same API has several mistakes and doesn't even attempt part of the text.

Can someone tell me why this is, am I missing some kind of setting?

** Also using android play services, gives incorrect results - It would be my preference to use play services

1
androidgoogle-vision

1 Answers

2
votes

I found a clue in this SO post - Turns out that if you change the API feature from TEXT_DETECTION to DOCUMENT_TEXT_DETECTION it works 100% accurately.

I'm going to leave this question up in case anyone else gets the same issues.