如何在 Temi Robot 上旋转活动物体检测屏幕？

如何解决如何在 Temi Robot 上旋转活动物体检测屏幕？

我目前正在使用 android-demo-app/ObjectDetection/ 在 Temi Robot 上，预加载的图像目前可以正常工作，但是当我按“实时”进入实时物体检测屏幕时，它向右旋转了 90 度。

Temi 机器人在屏幕的同一侧只有一个前置摄像头。

我尝试更改 textureView.setTransform() imageAnalysisConfig.Builder().setTargetRotation() imageAnalysis.setTargetRotation() 但无济于事

还尝试将 AndroidManifest.xml 标记下的 screenorientation activity 更改为 fullSenor 或 Landscape，但没有任何改变。

我一直在 Android Developer CameraX 页面上下寻找答案 first link second link，但我找不到任何答案。也许我不够聪明，无法在这里找到解决方案。

非常感谢任何帮助！

AbstactCameraXActivity.java

    private void setupCameraX() {
        final TextureView textureView = getCameraPreviewTextureView();
        final PreviewConfig previewConfig = new PreviewConfig.Builder().build();
        final Preview preview = new Preview(previewConfig);
//        Matrix m = new Matrix();
//        m.postRotate(180);
//        textureView.setTransform(m); //not working
        preview.setonPreviewOutputUpdateListener(output -> textureView.setSurfaceTexture(output.getSurfaceTexture()));



        final var imageAnalysisConfig =
                new ImageAnalysisConfig.Builder()
                        .setTargetResolution(new Size(500,500))
                        .setCallbackHandler(mBackgroundHandler)
                        .setimageReaderMode(ImageAnalysis.ImageReaderMode.ACQUIRE_LATEST_IMAGE)
                        //.setTargetRotation(Surface.ROTATION_0) // not working
                        .build();

         imageAnalysis = new ImageAnalysis(imageAnalysisConfig);
        imageAnalysis.setAnalyzer((image,rotationdegrees) -> {
            if (SystemClock.elapsedRealtime() - mLastAnalysisResultTime < 500) {
                return;
            }

            final R2 result = analyzeImage(image,rotationdegrees);
            if (result != null) {
                mLastAnalysisResultTime = SystemClock.elapsedRealtime();
                runOnUiThread(() -> applyToUiAnalyzeImageResult(result));
            }
        });
        //imageAnalysis.setTargetRotation(Surface.ROTATION_180); // not working
        CameraX.bindToLifecycle(this,preview,imageAnalysis);
    }

ObjectDetectionActivity.java

@Override
@WorkerThread
@Nullable
protected AnalysisResult analyzeImage(ImageProxy image,int rotationdegrees) {
    try {
        if (mModule == null) {
            mModule = LiteModuleLoader.load(MainActivity.assetFilePath(getApplicationContext(),"yolov5s.torchscript.ptl"));
        }
    } catch (IOException e) {
        Log.e("Object Detection","Error reading assets",e);
        return null;
    }

    Bitmap bitmap = imgToBitmap(Objects.requireNonNull(image.getimage()));
    Matrix matrix = new Matrix();
    matrix.postRotate(90.0f);
    bitmap = Bitmap.createBitmap(bitmap,bitmap.getWidth(),bitmap.getHeight(),matrix,true);
    Bitmap resizedBitmap = Bitmap.createScaledBitmap(bitmap,PrePostProcessor.mInputWidth,PrePostProcessor.mInputHeight,true);

    final Tensor inputTensor = TensorImageUtils.bitmapToFloat32Tensor(resizedBitmap,PrePostProcessor.NO_MEAN_RGB,PrePostProcessor.NO_STD_RGB);
    IValue[] outputTuple = mModule.forward(IValue.from(inputTensor)).toTuple();
    final Tensor outputTensor = outputTuple[0].toTensor();
    final float[] outputs = outputTensor.getDataAsFloatArray();

    float imgScaleX = (float)bitmap.getWidth() / PrePostProcessor.mInputWidth;
    float imgScaleY = (float)bitmap.getHeight() / PrePostProcessor.mInputHeight;
    float ivScaleX = (float)mResultView.getWidth() / bitmap.getWidth();
    float ivScaleY = (float)mResultView.getHeight() / bitmap.getHeight();

    final ArrayList<Result> results = PrePostProcessor.outputsToNMSPredictions(outputs,imgScaleX,imgScaleY,ivScaleX,ivScaleY,0);
    return new AnalysisResult(results);
}

AndroidManifest.xml

<?xml version="1.0" encoding="utf-8"?>
<manifest xmlns:android="http://schemas.android.com/apk/res/android"
    package="org.pytorch.demo.objectdetection">

    <uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE" />
    <uses-permission android:name="android.permission.CAMERA" />

    <application
        android:allowBackup="true"
        android:icon="@mipmap/ic_launcher"
        android:label="@string/app_name"
        android:roundIcon="@mipmap/ic_launcher_round"
        android:supportsRtl="true"
        android:theme="@style/AppTheme">
        <activity android:name=".MainActivity"
            android:configChanges="orientation"
            android:screenorientation="fullSensor">
            <intent-filter>
                <action android:name="android.intent.action.MAIN" />
                <category android:name="android.intent.category.LAUNCHER" />
            </intent-filter>
        </activity>
        <activity
            android:name=".ObjectDetectionActivity"
            android:configChanges="orientation"
            android:screenorientation="fullSensor">
        </activity>
    </application>

</manifest>

更新

我想我现在可能知道问题所在了。在 ObjectDetectionActivity 的 setupCameraX() 方法中，我应该操纵 textureView 并操纵矩阵变换的枢轴是我所需要的。我开始在屏幕上看到一些cameraview。但是我不知道这个参数中需要的 x 和 y 是什么...

      final TextureView textureView = getCameraPreviewTextureView();
    final PreviewConfig previewConfig = new PreviewConfig.Builder().build();
    final Preview preview = new Preview(previewConfig);
    Matrix m = new Matrix();
    m.postRotate(180,x,y);//potential solution here.
    textureView.setTransform(m); //not working
    preview.setonPreviewOutputUpdateListener(output -> textureView.setSurfaceTexture(output.getSurfaceTexture()));

解决方法

我已将 cameraX 版本从 1.0.0-alpha5 更改为 1.0.0

  private void setupCameraX() {
    ListenableFuture<ProcessCameraProvider> cameraProviderFuture =
            ProcessCameraProvider.getInstance(this);

    cameraProviderFuture.addListener(() -> {
        try {
            ProcessCameraProvider cameraProvider = cameraProviderFuture.get();
            PreviewView previewView = getCameraPreviewTextureView();
            final Preview preview = new Preview.Builder()
                    .setTargetRotation(Surface.ROTATION_270)//working nicely
                    .build();
            //TODO: Check if result_view can render over preview_view


            CameraSelector cameraSelector = new CameraSelector
                    .Builder()
                    .requireLensFacing(CameraSelector.LENS_FACING_FRONT)
                    .build();

            preview.setSurfaceProvider(previewView.getSurfaceProvider());

            executor = Executors.newSingleThreadExecutor();
            imageAnalysis = new ImageAnalysis.Builder()
                    .setTargetResolution(new Size(500,500))

                    .setBackpressureStrategy(ImageAnalysis.STRATEGY_KEEP_ONLY_LATEST)
                    .build();
            imageAnalysis.setAnalyzer(executor,image -> {
                        Log.d("image analyzer","Entered Analyse method");
                        if (SystemClock.elapsedRealtime() - mLastAnalysisResultTime < 500) {
                            return;
                        }

                        final T result = analyzeImage(image,90);
                        if (result != null) {
                            mLastAnalysisResultTime = SystemClock.elapsedRealtime();
                            runOnUiThread(() -> applyToUiAnalyzeImageResult(result));
                        }
                    });
            camera = cameraProvider.bindToLifecycle(
                    this,cameraSelector,imageAnalysis,preview);
        } catch (InterruptedException | ExecutionException e) {
            new AlertDialog
                    .Builder(this)
                    .setTitle("Camera setup error")
                    .setMessage(e.getMessage())
                    .setPositiveButton("Ok",(dialog,which) -> {
                            })
                    .show();
        }
    },ContextCompat.getMainExecutor(this));

注意：getCameraPreviewTextureView() 是膨胀 ViewStub 的方法。我只是在关注 pytorch android 示例。

  @Override
protected PreviewView getCameraPreviewTextureView() {
    mResultView = findViewById(R.id.resultView);
    //
    return ((ViewStub) findViewById(R.id.preview_view_stub))
            .inflate()
            .findViewById(R.id.preview_view);
}