Running a MobileNet on Android using Tensorflow Mobile

Setup

Getting started with TF Lite is very easy

1. First download the model. We will use MobileNet which can be downloaded here.
2. Place the model file in the assets folder, along with the associated labels for the model

    private MappedByteBuffer loadModelFile(AssetManager assetManager, String modelPath) throws IOException {
        AssetFileDescriptor fileDescriptor = assetManager.openFd(modelPath);
        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);
    }
   

3. Load the model as a TensorFlow Interpreter

   interpreter = new Interpreter(classifier.loadModelFile(assetManager, modelPath));
   

4. Use the model

   public List<Recognition> recognizeImage(Bitmap bitmap) {
        ByteBuffer byteBuffer = convertBitmapToByteBuffer(bitmap);
        byte[][] result = new byte[1][labelList.size()];
        interpreter.run(byteBuffer, result);
        return getSortedResult(result);
    }