[Tensorflow] TensorFlow Lite Tutorial

一个自定义demo

一、学习资料

Ref: Tensorflow lite 系列视频集，学习TensorFlow Lite的部署实践。[有点过时]

Official sample: TensorFlow Lite example apps

Object Detection: android/iso --> https://github.com/tensorflow/examples/tree/master/lite/examples/object_detection

 

 

二、代码分析 

 

tflite定义了分类器的接口和具体的一种实现。

  @Override
  public List<Recognition> recognizeImage(final Bitmap bitmap)

 

DetectorActivity.java继承了CameraActivity.java，在其中调用 recognizeImage()。

  @Override
  protected void processImage() {

    ++timestamp;
    final long currTimestamp = timestamp;
    trackingOverlay.postInvalidate();

    // No mutex needed as this method is not reentrant.
    if (computingDetection) {
      readyForNextImage();
      return;
    }
    computingDetection = true;
    LOGGER.i("Preparing image " + currTimestamp + " for detection in bg thread.");

    rgbFrameBitmap.setPixels(getRgbBytes(), 0, previewWidth, 0, 0, previewWidth, previewHeight);

    readyForNextImage();

    final Canvas canvas = new Canvas(croppedBitmap);
    canvas.drawBitmap(rgbFrameBitmap, frameToCropTransform, null);
    // For examining the actual TF input.
    if (SAVE_PREVIEW_BITMAP) {
      ImageUtils.saveBitmap(croppedBitmap);
    }

    runInBackground(
        new Runnable() {
          @Override
          public void run() {
            LOGGER.i("Running detection on image " + currTimestamp);
            final long startTime = SystemClock.uptimeMillis();
            /**
             * predict!
             */
            final List<Classifier.Recognition> results = detector.recognizeImage(croppedBitmap);
            lastProcessingTimeMs = SystemClock.uptimeMillis() - startTime;

            //--------------------------------------------------
            cropCopyBitmap = Bitmap.createBitmap(croppedBitmap);
            final Canvas canvas = new Canvas(cropCopyBitmap);
            final Paint paint = new Paint();
            paint.setColor(Color.RED);
            paint.setStyle(Style.STROKE);
            paint.setStrokeWidth(2.0f);
            //--------------------------------------------------

            float minimumConfidence = MINIMUM_CONFIDENCE_TF_OD_API;
            switch (MODE) {
              case TF_OD_API:
                minimumConfidence = MINIMUM_CONFIDENCE_TF_OD_API;
                break;
            }

            final List<Classifier.Recognition> mappedRecognitions =
                new LinkedList<Classifier.Recognition>();

            for (final Classifier.Recognition result : results) {
              final RectF location = result.getLocation();
              if (location != null && result.getConfidence() >= minimumConfidence) {

                canvas.drawRect(location, paint);
                cropToFrameTransform.mapRect(location);

                result.setLocation(location);
                mappedRecognitions.add(result);
              }
            }

            ////////////////////////////////////////////////////////
            // implement, call jni opencv
            ////////////////////////////////////////////////////////

            tracker.trackResults(mappedRecognitions, currTimestamp);
            trackingOverlay.postInvalidate();

            computingDetection = false;

            runOnUiThread(
                new Runnable() {
                  @Override
                  public void run() {
                    showFrameInfo(previewWidth + "x" + previewHeight);
                    showCropInfo(cropCopyBitmap.getWidth() + "x" + cropCopyBitmap.getHeight());
                    showInference(lastProcessingTimeMs + "ms");
                  }
                });
          }
        });
  }

 

 

 

模型转换

一、识别模型

在assets文件夹中。

 

 

 

/* implement */