TF Lite in Google Play services has stable ABI version 1.11.0 which is less than the required version 1.13.0

[rachitjoshi9]

We are using mobile_scanner: ^3.5.1 version in our flutter project, on android device when ever camera opens below warning keeps on coming repeatedly

W/native (30072): W0000 00:00:1698456375.622782 32103 abi_method_util.h:33] not supported: TfLiteInterpreterGetTensor: TF Lite in Google Play services has stable ABI version 1.11.0 which is less than the required version 1.13.0.

environment: sdk: '>=2.12.0 <3.0.0'

Where do we need to make changes to upgrade to required version 1.13.0

[navaronbracke]

As far as I know, Tensor Flow Lite is a dependency of MLKit. I think we have to bump the version of MLKit that is used to get rid of this warning. I'll look into this.

[rachitjoshi9]

Thanks for update, we will be waiting for this change to be done in new version.

[navaronbracke]

I updated MLKit on Android to version 18.3.0, which is the latest available version. Not sure if this will fix it, since the changelog doesn't mention this bug.

Will publish a new version with the updated MLKit today.

[dhirajmishra98]

the issue is still pending!

[navaronbracke]

@dhirajmishra98 As I mentioned in https://github.com/juliansteenbakker/mobile_scanner/issues/836#issuecomment-1787248290 the MLKit version has been updated to the latest available version in mobile_scanner 3.5.2. If this issue is still occurring, we cannot do anything at the moment since version 18.3.0 is the latest version.

See https://developers.google.com/ml-kit/release-notes#latest_versions

[navaronbracke]

Apparently support for version 1.13.0 is still in preview at TensorFlow. Is anyone using preview/beta versions here?

[navaronbracke]

Since both Tensorflow and MLKit coordinate releases, we'll have to wait until the fix in Tensorflow is no longer in preview. Only then will MLKit be able to create a new release.

Once that happens, I'll bump our MLKit version here.

[LaPyaeOo]

This issue still happen. I used mobile_scanner: ^3.5.2. After that first time opening the camera still get

TF Lite in Google Play services has stable ABI version 1.11.0 which is less than the required version 1.13.0

and can't use. Only work in second time after enter again to this screen.

[asavchuk]

I use simple_barcode_scanner and got a lot of outputs in console

TF Lite in Google Play services has stable ABI version 1.11.0 which is less than the required version 1.13.0.

[YedekTunahan]

TF Lite on Google Play services has a stable ABI version of 1.13.0, which is less than the required version 1.11.0. Does the problem still persist implementation 'com.google.android.gms:play-services-mlkit-text-recognition:19.0.0' version

[officialismailshah]

yes problem still persists...

[officialismailshah]

is there a work around it?

[officialismailshah]

For beginners, I have used the bundled version, there is no exception about the TF Lite in Google Play services has stable ABI version 1.11.0 which is less than the required version 1.13.0 just remove "dev.steenbakker.mobile_scanner.useUnbundled=true"

[iruuuuull]

Is there any way to solve it?

[navaronbracke]

For anyone still waiting for a solution: We need to wait until there is a new release listed in https://developers.google.com/ml-kit/release-notes#android-api-latest-versions

Otherwise I cannot fix this issue. In the meantime I suggest using the unbundled version, as mentioned in https://github.com/juliansteenbakker/mobile_scanner/issues/836#issuecomment-1826439802

[fergushenderson]

I'm on the "TF Lite in Play services" team at Google, and I just came across this issue. (For faster support on similar issues in future, file an issue in the tensorflow github issues list, and tag it with the appropriate tags to indicate that it is related to TF Lite in Play services.)

I believe this issue should be fixed already, by an updated version of TF Lite in Play services, with no need for code changes or even recompilation for apps that were previously affected. If the issue is still occuring, please let us know.

[navaronbracke]

Thanks for the update

I'm on the "TF Lite in Play services" team at Google, and I just came across this issue. (For faster support on similar issues in future, file an issue in the tensorflow github issues list, and tag it with the appropriate tags to indicate that it is related to TF Lite in Play services.)

I believe this issue should be fixed already, by an updated version of TF Lite in Play services, with no need for code changes or even recompilation for apps that were previously affected. If the issue is still occuring, please let us know.

Thanks for the update. I'll keep that in mind for future similar reports here.