Repository navigation
On Device Recognition not accuracy on Android, iOS work perfect #21
Description
I'm developing an on-device recognition app that works offline both Android an iOS.
I've tested on Android (Galaxy Note 9, Nokia 8.1, Galaxy S9, Galaxy Note 5): all of the phones seem hard to recognize text, 10 times tried, 1 time detected. It's really hard to work on Android
For iPhone (iPhone 5s, iPhone 6s, iPhone X): all of the phones detect like 99%. I've tried 10 times and all 10 times it shows correct 100%.
I use the code on Example.
Is there any step I need to do on Android? Sine on this document https://firebase.google.com/docs/ml-kit/android/recognize-text (Before you begin, number 3) it's said that the model will automatically download on the first time we use on-device text recognition
Thank you.
Activity
maximmcnair commentedon Oct 29, 2019
@tienlx91 Did you manage to work around this?
tayguohong commentedon Feb 26, 2020
@tienlx91 i have the issue totally contradict with yours. Android platform works well, but really struggle to recognise text in iPhone 7, do you found any solution on this issue?
tienlx91 commentedon Mar 4, 2020
For the first time we use the recognizer, it will automatically download the dictionary to the device. So I have to wait a little bit to test again, but the accuracy on Android still low (3/10)
So I switched to iOS only, worked perfect
on the iPhone I need to do 1 more step, since the image very high resolution, I use an image convert to lower the resolution
tayguohong commentedon Mar 4, 2020
In IOS is image resolution issue? I have tested on iOS device last week, on portrait orientation is almost impossible to recognise text. But it works on landscape orientation. I suspect there is orientation issue on iOS platform when the
fixOrientation: trueis set in options. Have you encountered this issue?tuledev commentedon May 26, 2020
@tayguohong I got the same issue. Using
forceUpOrientation: trueontakePictureAsyncwill fix the problem