object_detection_retina_a311d.pbtxt

# Images on CPU coming into and out of the graph.
input_stream: "input_frame"
output_stream: "output_detect"

# Transforms the input image on CPU to a 320x320 image. To scale the image, by
# default it uses the STRETCH scale mode that maps the entire input image to the
# entire transformed image. As a result, image aspect ratio may be changed and
# objects in the image may be deformed (stretched or squeezed), but the object
# detection model used in this graph is agnostic to that deformation.
node {
  calculator: "ImageTransformationCalculator"
  input_stream: "IMAGE:input_frame"
  output_stream: "IMAGE:transformed_input_frame"
  node_options: {
    [type.googleapis.com/mediapipe.ImageTransformationCalculatorOptions] {
      output_width: 960
      output_height: 544
    }
  }
}

# Converts the transformed input image on CPU into an image tensor stored as a
# TfLiteTensor.
node {
  calculator: "TengineConverterCalculator"
  input_stream: "IMAGE:transformed_input_frame"
  output_stream: "ARRAYS:image_tensor"
  node_options: {
    [type.googleapis.com/mediapipe.TengineConverterCalculatorOptions] {
      tensor_mean: {val1:104 val2:117 val3:123}
      tensor_scale: {val1:1 val2:1 val3:1}
    }
  }
}

node {
  calculator: "TengineInferenceCalculator"
  input_stream: "ARRAYS:image_tensor"
  output_stream: "ARRAYS:detection_tensors"
  output_stream: "TENSOR_SHAPE:tensor_shapes"
  output_stream: "QUANT_PARAM:quant_param"
  node_options: {
    [type.googleapis.com/mediapipe.TengineInferenceCalculatorOptions] {
      model_path: "../models/retinaface_multi_uint8.tmfile"
      data_type: "uint8"
      output_num: 2
      max_dim: 4
      tengine_backend: "timvx"
    }
  }
}


node {
  calculator: "RetinaAnchorsCalculator"
  output_side_packet: "anchors"
  node_options: {
    [type.googleapis.com/mediapipe.RetinaAnchorsCalculatorOptions] {
      num_layers: 3
      min_scale: 0.2
      max_scale: 0.95
      input_size_height: 544
      input_size_width: 960
      anchor_offset_x: 0.5
      anchor_offset_y: 0.5
      strides: 8
      strides: 16
      strides: 32
      aspect_ratios: 1.0
      aspect_ratios: 2.5
      reduce_boxes_in_lowest_layer: false
    }
  }
}


node {
  calculator: "TengineTensorsToDetectionsCalculator"
  input_stream: "ARRAYS:detection_tensors"
  input_side_packet: "ANCHORS:anchors"
  input_stream: "QUANT_PARAM:quant_param"
  output_stream: "DETECTIONS:detections"
  node_options: {
    [type.googleapis.com/mediapipe.TengineTensorsToDetectionsCalculatorOptions] {
      num_classes: 4
      num_boxes: 85680
      num_coords: 4
      min_score_thresh: 0.5
      sigmoid_score: true
      data_type: "uint8"
    }
  }
}

# Performs non-max suppression to remove excessive detections.
node {
  calculator: "NonMaxSuppressionCalculator"
  input_stream: "detections"
  output_stream: "output_detect"
  node_options: {
    [type.googleapis.com/mediapipe.NonMaxSuppressionCalculatorOptions] {
      min_suppression_threshold: 0.4
      # max_num_detections: 100
      overlap_type: INTERSECTION_OVER_UNION
      return_empty_detections: true
    }
  }
}