Android CameraX stuck with two use cases

Question

I'm novice in Android development (more Python and ML engineer) but wanted to try this example from TensorFlow: TF Lite Transfer Learning.

I succesfully run it on Android Studio but spotted that I cannot do anything with the app as it works extraordinary slow. I was digging through the code to find a root cause a found out this.

In function CameraFragment::startCamera() a preview context is created

    PreviewConfig config = new PreviewConfig.Builder()
    .setLensFacing(LENS_FACING)
    .setTargetAspectRatio(screenAspectRatio)
    .setTargetRotation(viewFinder.getDisplay().getRotation())
    .build();

Preview preview = new Preview(config);

preview.setOnPreviewOutputUpdateListener(previewOutput -> {
  ViewGroup parent = (ViewGroup) viewFinder.getParent();
  parent.removeView(viewFinder);
  parent.addView(viewFinder, 0);

As well as other use case that we can keep empty:

    final ImageAnalysisConfig imageAnalysisConfig =
        new ImageAnalysisConfig.Builder()
                .setLensFacing(LENS_FACING)
                .setTargetResolution(new Size(224, 224))
                .setCallbackHandler(new Handler(inferenceThread.getLooper()))
                .setImageReaderMode(ImageAnalysis.ImageReaderMode.ACQUIRE_LATEST_IMAGE)
                .build();
final ImageAnalysis imageAnalysis2 = new ImageAnalysis(imageAnalysisConfig);
imageAnalysis2.setAnalyzer((image, rotationDegrees) -> { });

New we have a line to bring them to life:

CameraX.bindToLifecycle(this, preview, imageAnalysis2);

And here is where the problem starts. If we keep them like that the application is unusable. But if we keep only one (either one) we can use the app to the point that all novigation works normally. Do you know what is causing this behaviour?

Answer 1

This build.gradle uses a rather old version; migrate to version 1.0.0-beta01.