后台协程android如何处理TensorImage

Question

我正在尝试在后台处理图像以检测来自相机图像的对象，但是当我尝试在后台协程中处理它们时，它会出错。

java.lang.IllegalStateException: Error occurred when detecting the image: Invoke() failed.
at org.tensorflow.lite.task.vision.detector.ObjectDetector.detectNative(Native Method)                                                                                              at org.tensorflow.lite.task.vision.detector.ObjectDetector.detect(ObjectDetector.java:528)
                                                                                                        at org.tensorflow.lite.task.vision.detector.ObjectDetector.access$1200(ObjectDetector.java:88)
at org.tensorflow.lite.task.vision.detector.ObjectDetector$4.run(ObjectDetector.java:481)
at org.tensorflow.lite.task.vision.detector.ObjectDetector$4.run(ObjectDetector.java:477)
at org.tensorflow.lite.task.vision.core.BaseVisionTaskApi.run(BaseVisionTaskApi.java:49)
at org.tensorflow.lite.task.vision.detector.ObjectDetector.detect(ObjectDetector.java:476)
at org.tensorflow.lite.task.vision.detector.ObjectDetector.detect(ObjectDetector.java:446)
at com.nst.n692_android.ObjectDetectorHelper$detectAsync$2.invokeSuspend(ObjectDetectorHelper.kt:167)
at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:33)
at kotlinx.coroutines.DispatchedTask.run(DispatchedTask.kt:106)
at kotlinx.coroutines.internal.LimitedDispatcher.run(LimitedDispatcher.kt:42)
at kotlinx.coroutines.scheduling.TaskImpl.run(Tasks.kt:95)
at kotlinx.coroutines.scheduling.CoroutineScheduler.runSafely(CoroutineScheduler.kt:570)
at kotlinx.coroutines.scheduling.CoroutineScheduler$Worker.executeTask(CoroutineScheduler.kt:750)
at kotlinx.coroutines.scheduling.CoroutineScheduler$Worker.runWorker(CoroutineScheduler.kt:677)
at kotlinx.coroutines.scheduling.CoroutineScheduler$Worker.run(CoroutineScheduler.kt:664)                                                                                                       Suppressed: kotlinx.coroutines.DiagnosticCoroutineContextException: [StandaloneCoroutine{Cancelling}@7613710, Dispatchers.Main.immediate]

我的代码是这样的：

return withContext(Dispatchers.IO){
            // Preprocess the image and convert it into a TensorImage for detection.

             val inputImage = TensorImage.fromBitmap(rotatedImage)
            val outputs = objectDetector?.detect(inputImage)
            if (outputs.isNullOrEmpty())
                return@withContext null
            val detectionResult = outputs[0]
            val category = detectionResult?.boundingBox ?: return@withContext null
            val cordLeft =
                PointF(category.left, category.top)

            val cordRight =
                PointF(category.right, category.top)
            inferenceTime = SystemClock.uptimeMillis() - inferenceTime
            val rotatedCoordinatesLeft =
                cordLeft.rotateCoordinates(rotatedImage.width, rotatedImage.height, imageRotation)
            val rotatedCoordinatesRight =
                cordRight.rotateCoordinates(rotatedImage.width, rotatedImage.height, imageRotation)
             return@withContext DetectedObjectResult(
                detectionResult.categories[0].score.toString(),
                rotatedCoordinatesLeft,
                rotatedCoordinatesRight
            )
        }