0

this is more a question for the Firebase/MLKit team.

When using the same image as input for the iOS and Android versions of MLKit Text Recognition On-Device API, I do get different results on the bounding boxes information (x,y,width,height) provided by each solution.

Please see below the original input image I've used for my tests and another image that shows the resulting text block's bounding boxes that were drawn based on the information provided by both of the Text Recognition on-device APIs (in blue is the iOS result and in red is the Android one):

Does anyone know what could cause such differences between the results for iOS and Android versions of the API? I suspect they use different ML models for the text recognition / extraction of bounding boxes. If so, is there any chance of having both solutions running the same model in a near future since they are still on beta release?

Any thoughts are welcome!

Joao
  • 1
  • I admit I have never used the MLKit API. Is the actual image analyzed or the display view. If it is the display view (like `ImageView`) than I could image this being a display resolution issue. – Barns Aug 08 '18 at 14:57
  • I don't think this is a display resolution issue. I've used the actual image as input for the text recognition process on both platform. Basically, I've appended both results into a file and used this data to drawn the boxes on the result image. – Joao Aug 08 '18 at 15:12

1 Answers1

1

Your are right. The underlying engines for iOS and Android are different in ML Kit for now. We will update the models to make them consistent in later releases.

Shiyu
  • 875
  • 4
  • 5
  • Any update on the models? I am currently facing the same issue on my production environment. Is there any way to track this somewhere in firebase repo or something? – Aayush Thakur Sep 13 '18 at 09:30