删除video检测的代码，添加延时输出的代码。

app/android/java/com/khronos/vulkan_samples/FaceLandmarkerHelper.kt

@@ -139,6 +139,51 @@ class FaceLandmarkerHelper(
         }
     }
 
+    private var frameId: Long = 0
+    // 简化的延时跟踪
+    private val frameTimings = mutableMapOf<Long, FrameTiming>()
+    private var lastFpsUpdateTime: Long = 0
+    private var frameCount: Int = 0
+    private var currentFps: Double = 0.0
+
+    // 内部数据类用于存储时间信息
+    private data class FrameTiming(
+        var captureTime: Long = 0,
+        var preprocessStartTime: Long = 0,
+        var preprocessEndTime: Long = 0,
+        var detectionStartTime: Long = 0
+    )
+
+    // 辅助方法：记录帧时间信息
+    private fun recordFrameTiming(frameId: Long, block: (FrameTiming) -> Unit) {
+        val timing = frameTimings.getOrPut(frameId) { FrameTiming() }
+        block(timing)
+    }
+
+    // 辅助方法：更新FPS计算
+    private fun updateFps() {
+        frameCount++
+        val currentTime = SystemClock.uptimeMillis()
+
+        if (lastFpsUpdateTime == 0L) {
+            lastFpsUpdateTime = currentTime
+            return
+        }
+
+        val timeDiff = currentTime - lastFpsUpdateTime
+        if (timeDiff >= 1000) { // 每秒更新一次FPS
+            currentFps = frameCount * 1000.0 / timeDiff
+            frameCount = 0
+            lastFpsUpdateTime = currentTime
+        }
+    }
+
+    // 辅助方法：清理旧的时间记录
+    private fun cleanupOldTimings(currentFrameId: Long) {
+        val framesToRemove = frameTimings.keys.filter { it < currentFrameId - 10 }
+        framesToRemove.forEach { frameTimings.remove(it) }
+    }
+
     // Convert the ImageProxy to MP Image and feed it to FacelandmakerHelper.
     fun detectLiveStream(
         imageProxy: ImageProxy,
@@ -152,6 +197,15 @@ class FaceLandmarkerHelper(
         }
         val frameTime = SystemClock.uptimeMillis()
 
+        val currentFrameId = frameId++
+        val captureTime = SystemClock.uptimeMillis()
+
+        // 记录捕获时间
+        recordFrameTiming(currentFrameId) {
+            it.captureTime = captureTime
+            it.preprocessStartTime = SystemClock.uptimeMillis()
+        }
+
         // Copy out RGB bits from the frame to a bitmap buffer
         val bitmapBuffer =
             Bitmap.createBitmap(
@@ -198,152 +252,54 @@ class FaceLandmarkerHelper(
         // be returned in returnLivestreamResult function
     }
 
-    // Accepts the URI for a video file loaded from the user's gallery and attempts to run
-    // face landmarker inference on the video. This process will evaluate every
-    // frame in the video and attach the results to a bundle that will be
-    // returned.
-    fun detectVideoFile(
-        videoUri: Uri,
-        inferenceIntervalMs: Long
-    ): VideoResultBundle? {
-        if (runningMode != RunningMode.VIDEO) {
-            throw IllegalArgumentException(
-                "Attempting to call detectVideoFile" +
-                        " while not using RunningMode.VIDEO"
-            )
-        }
 
-        // Inference time is the difference between the system time at the start and finish of the
-        // process
-        val startTime = SystemClock.uptimeMillis()
-
-        var didErrorOccurred = false
-
-        // Load frames from the video and run the face landmarker.
-        val retriever = MediaMetadataRetriever()
-        retriever.setDataSource(context, videoUri)
-        val videoLengthMs =
-            retriever.extractMetadata(MediaMetadataRetriever.METADATA_KEY_DURATION)
-                ?.toLong()
-
-        // Note: We need to read width/height from frame instead of getting the width/height
-        // of the video directly because MediaRetriever returns frames that are smaller than the
-        // actual dimension of the video file.
-        val firstFrame = retriever.getFrameAtTime(0)
-        val width = firstFrame?.width
-        val height = firstFrame?.height
-
-        // If the video is invalid, returns a null detection result
-        if ((videoLengthMs == null) || (width == null) || (height == null)) return null
-
-        // Next, we'll get one frame every frameInterval ms, then run detection on these frames.
-        val resultList = mutableListOf<FaceLandmarkerResult>()
-        val numberOfFrameToRead = videoLengthMs.div(inferenceIntervalMs)
-
-        for (i in 0..numberOfFrameToRead) {
-            val timestampMs = i * inferenceIntervalMs // ms
-
-            retriever
-                .getFrameAtTime(
-                    timestampMs * 1000, // convert from ms to micro-s
-                    MediaMetadataRetriever.OPTION_CLOSEST
-                )
-                ?.let { frame ->
-                    // Convert the video frame to ARGB_8888 which is required by the MediaPipe
-                    val argb8888Frame =
-                        if (frame.config == Bitmap.Config.ARGB_8888) frame
-                        else frame.copy(Bitmap.Config.ARGB_8888, false)
-
-                    // Convert the input Bitmap object to an MPImage object to run inference
-                    val mpImage = BitmapImageBuilder(argb8888Frame).build()
-
-                    // Run face landmarker using MediaPipe Face Landmarker API
-                    faceLandmarker?.detectForVideo(mpImage, timestampMs)
-                        ?.let { detectionResult ->
-                            resultList.add(detectionResult)
-                        } ?: {
-                        didErrorOccurred = true
-                        faceLandmarkerHelperListener?.onError(
-                            "ResultBundle could not be returned" +
-                                    " in detectVideoFile"
-                        )
-                    }
-                }
-                ?: run {
-                    didErrorOccurred = true
-                    faceLandmarkerHelperListener?.onError(
-                        "Frame at specified time could not be" +
-                                " retrieved when detecting in video."
-                    )
-                }
-        }
-
-        retriever.release()
-
-        val inferenceTimePerFrameMs =
-            (SystemClock.uptimeMillis() - startTime).div(numberOfFrameToRead)
-
-        return if (didErrorOccurred) {
-            null
-        } else {
-            VideoResultBundle(resultList, inferenceTimePerFrameMs, height, width)
-        }
-    }
-
-    // Accepted a Bitmap and runs face landmarker inference on it to return
-    // results back to the caller
-    fun detectImage(image: Bitmap): ResultBundle? {
-        if (runningMode != RunningMode.IMAGE) {
-            throw IllegalArgumentException(
-                "Attempting to call detectImage" +
-                        " while not using RunningMode.IMAGE"
-            )
-        }
-
-
-        // Inference time is the difference between the system time at the
-        // start and finish of the process
-        val startTime = SystemClock.uptimeMillis()
-
-        // Convert the input Bitmap object to an MPImage object to run inference
-        val mpImage = BitmapImageBuilder(image).build()
-
-        // Run face landmarker using MediaPipe Face Landmarker API
-        faceLandmarker?.detect(mpImage)?.also { landmarkResult ->
-            val inferenceTimeMs = SystemClock.uptimeMillis() - startTime
-            return ResultBundle(
-                landmarkResult,
-                inferenceTimeMs,
-                image.height,
-                image.width
-            )
-        }
-
-        // If faceLandmarker?.detect() returns null, this is likely an error. Returning null
-        // to indicate this.
-        faceLandmarkerHelperListener?.onError(
-            "Face Landmarker failed to detect."
-        )
-        return null
-    }
 
     // Return the landmark result to this FaceLandmarkerHelper's caller
     private fun returnLivestreamResult(
         result: FaceLandmarkerResult,
         input: MPImage
     ) {
+        val resultTime = SystemClock.uptimeMillis()
+        val frameId = result.timestampMs()
         if( result.faceLandmarks().size > 0 ) {
-            val finishTimeMs = SystemClock.uptimeMillis()
+            // 计算各种延时
-            val inferenceTime = finishTimeMs - result.timestampMs()
+            val timings = frameTimings[frameId]
+            if (timings != null) {
+                val totalLatency = resultTime - timings.captureTime
+                val preprocessLatency = timings.preprocessEndTime - timings.preprocessStartTime
+                val detectionLatency = resultTime - timings.detectionStartTime
+                val captureToDetectionLatency = timings.detectionStartTime - timings.captureTime
 
-            faceLandmarkerHelperListener?.onResults(
+                Log.d("TimeLatency",
-                ResultBundle(
+                    "总延时: ${totalLatency}ms | " +
-                    result,
+                            "预处理: ${preprocessLatency}ms | " +
-                    inferenceTime,
+                            "检测: ${detectionLatency}ms | " +
-                    input.height,
+                            "捕获到检测: ${captureToDetectionLatency}ms | " +
-                    input.width
+                            "FPS: ${String.format("%.1f", currentFps)}"
                 )
-            )
+
+                val finishTimeMs = SystemClock.uptimeMillis()
+                val inferenceTime = finishTimeMs - result.timestampMs()
+                faceLandmarkerHelperListener?.onResults(
+                    ResultBundle(
+                        result,
+                        inferenceTime, // 主要是检测时间
+                        input.height,
+                        input.width,
+                    )
+                )
+            } else {
+                val finishTimeMs = SystemClock.uptimeMillis()
+                val inferenceTime = finishTimeMs - result.timestampMs()
+                faceLandmarkerHelperListener?.onResults(
+                    ResultBundle(
+                        result,
+                        inferenceTime,
+                        input.height,
+                        input.width
+                    )
+                )
+            }
         }
         else {
             faceLandmarkerHelperListener?.onEmpty()
@@ -379,13 +335,6 @@ class FaceLandmarkerHelper(
         val inputImageWidth: Int,
     )
 
-    data class VideoResultBundle(
-        val results: List<FaceLandmarkerResult>,
-        val inferenceTime: Long,
-        val inputImageHeight: Int,
-        val inputImageWidth: Int,
-    )
-
     interface LandmarkerListener {
         fun onError(error: String, errorCode: Int = OTHER_ERROR)
         fun onResults(resultBundle: ResultBundle)

app/android/java/com/khronos/vulkan_samples/MainActivity.java

@@ -6,6 +6,7 @@ import android.content.pm.PackageManager;
 import android.os.Bundle;
 import android.os.Handler;
 import android.os.Looper;
+import android.os.SystemClock;
 import android.util.Log;
 import android.view.WindowManager;
 import android.widget.Toast;
@@ -182,6 +183,12 @@ public class MainActivity extends GameActivity implements FaceLandmarkerHelper.L
                     @Override
                     public void analyze(@NonNull ImageProxy image) {
                         try {
+                            long analyzeTime = SystemClock.elapsedRealtimeNanos();
+                            long captureTime = image.getImageInfo().getTimestamp();
+                            long latencyNs = analyzeTime - captureTime;
+                            double latencyMs = latencyNs / 1_000_000.0;
+
+                            Log.d("CameraLatency", String.format("Image capture to analyze latency: %.2f ms", latencyMs));
                             //Log.d("Camera", "Received image: " + image.getWidth() + "x" + image.getHeight());
                             processImageForVulkan(image);
                             detectFace(image);
@@ -315,7 +322,7 @@ public class MainActivity extends GameActivity implements FaceLandmarkerHelper.L
             nativeBuffer.order(ByteOrder.nativeOrder());
         }
 
-
+        long start_time =  System.currentTimeMillis();
 
 //        ArrayList<Vector> vertexs = new ArrayList<>();
 //        ArrayList<Integer> triangle_index = new ArrayList<>();
@@ -358,7 +365,10 @@ public class MainActivity extends GameActivity implements FaceLandmarkerHelper.L
         FloatBuffer floatBuffer = nativeBuffer.asFloatBuffer();
         floatBuffer.put(points);
         floatBuffer.position(0);
+        long cur_time =  System.currentTimeMillis();
+        Log.i("TimeLatency","Result Data ProcessTime:" + (cur_time-start_time));
         passDataToNative(nativeBuffer, index, width, height);
+        Log.i("TimeLatency","passDataToNative ProcessTime:" + (System.currentTimeMillis() - cur_time));
     }
 
     public float z_rate = -1.0f;