Fail to run my own sample after build. about tf-lite-unity-sample

[asus4]

The error shows that it failed to load native plugins libtensorflowlite_c.so and libtensorflowlite_gpu_delegate.so. The native android plugins are built for only ARM64 CPU for now. It doesn't support ARMv7 refs #11

Or it might also help to test if Posenet model works on the official Android sample.
https://github.com/tensorflow/examples/tree/master/lite/examples/posenet/android

[guoguoguilai]

The error shows that it failed to load native plugins libtensorflowlite_c.so and libtensorflowlite_gpu_delegate.so. The native android plugins are built for only ARM64 CPU for now. It doesn't support ARMv7 refs #11

Or it might also help to test if Posenet model works on the official Android sample.
https://github.com/tensorflow/examples/tree/master/lite/examples/posenet/android

Thanks for your quick response!
The CPU of my Android phone is ARM64-v8a. And I am also wondering why the sample from this repo works well in my phone even though throwed such error.

[guoguoguilai]

I am new to Tflite. I might make a stupid mistake. The privious tflite model cames from pb without quantization. After converting my pb model with a quantization option, the second error disappears. But the first error remains, which is strange.

[asus4]

sorry, I was misunderstood. In that case, I think the generated tflite model is not supported. If the model works on the official Android sample, you might want to try updating *.so library.

[guoguoguilai]

sorry, I was misunderstood. In that case, I think the generated tflite model is not supported. If the model works on the official Android sample, you might want to try updating *.so library.

But why this sample works well even though throw that error？
Maybe every time this error just occurs. The reason why nobody noticed that is no one look into adb log because their models work just fine.

[kaiquesacchi]

#77
I had a similar problem, the solution provided by @ganaware might help you:

I have had a similer problem.
The cause of the problem is that Mono is the default scripting backend for a new unity project for Android.
Unfortunately, Mono can run only for the ARMv7 architecture.

So my solution is that:

Project Settings > Player > Android > Other Settings > Configuration
Select IL2CPP for Scripting Backend
Uncheck ARMv7 and check ARM64 for Target Architectures
Or, build my own TFLite library for ARMv7
libnnapi_delegate.so, libtensorflowlite_c.so, libtensorflowlite_gpu_delegate.so
They worked fine for me.