Android Tensorflow Lite Interpreter crashes

Question

I'm loading a TensorFlow Lite model into my Android app, but the application keeps crashing (without any error printed out) on:

tflite = new Interpreter(loadModelFile(activity));

It is my understanding, from some issues on the tensorflow repo on github, that this is due to a bad tflite model; indeed, if I load a demo model from the repository instead, it works fine.

How do I check the correctness of a .tflite model? I created it from a tensorflow model file .pb converted with toco from the repo, which gave me neither errors nor positive feedback. The tensorflow model comes from a convertion of a keras model.

Can you look at the logcat and grab the stack trace around the crash? What kind of model are you using? Can you provide code snippet of your Android app? — Pannag Sanketi
Also you can try testing on python using the python interpreter if that is simpler for you: tensorflow.org/api_docs/python/tf/contrib/lite/Interpreter — suharshs

IbrahimOmar IbrahimOmar · Accepted Answer · 2020-04-20T05:12:51

You could use it as follows,[assuming your model is a file in the assets folder]

try{
    tflite = new Interpreter(loadModelFile());
} catch (Exception ex){
    ex.printStackTrace();
}


private MappedByteBuffer loadModelFile() throws IOException {
    AssetFileDescriptor fileDescriptor = this.getAssets().openFd("modelname.tflite");
    FileInputStream inputStream = new FileInputStream(fileDescriptor.getFileDescriptor());
    FileChannel fileChannel = inputStream.getChannel();
    long startOffset = fileDescriptor.getStartOffset();
    long declaredLength = fileDescriptor.getDeclaredLength();
    return fileChannel.map(FileChannel.MapMode.READ_ONLY, startOffset, declaredLength);
}

Android Tensorflow Lite Interpreter crashes

1 Answers