chromium/third_party/mediapipe/src/mediapipe/modules/objectron/objectron_detection_1stage_gpu.pbtxt

# MediaPipe Objectron detection gpu subgraph

type: "ObjectronDetectionSubgraphGpu"

input_stream: "IMAGE_GPU:input_video"
output_stream: "ANNOTATIONS:objects"

# Transforms the input image on GPU to a 480x640 image. To scale the input
# image, the scale_mode option is set to FIT to preserve the aspect ratio,
# resulting in potential letterboxing in the transformed image.
node: {
  calculator: "ImageTransformationCalculator"
  input_stream: "IMAGE_GPU:input_video"
  output_stream: "IMAGE_GPU:transformed_input_video"
  options: {
    [mediapipe.ImageTransformationCalculatorOptions.ext] {
      output_width: 480
      output_height: 640
      scale_mode: FIT
    }
  }
}

# Converts the transformed input image on GPU into an image tensor stored as a
# TfLiteTensor.
node {
  calculator: "TfLiteConverterCalculator"
  input_stream: "IMAGE_GPU:transformed_input_video"
  output_stream: "TENSORS_GPU:image_tensor"
}

# Generates a single side packet containing a TensorFlow Lite op resolver that
# supports custom ops needed by the model used in this graph.
node {
  calculator: "TfLiteCustomOpResolverCalculator"
  output_side_packet: "opresolver"
  options: {
    [mediapipe.TfLiteCustomOpResolverCalculatorOptions.ext] {
      use_gpu: true
    }
  }
}

# Runs a TensorFlow Lite model on GPU that takes an image tensor and outputs a
# vector of tensors representing, for instance, detection boxes/keypoints and
# scores.
node {
  calculator: "TfLiteInferenceCalculator"
  input_stream: "TENSORS_GPU:image_tensor"
  output_stream: "TENSORS:detection_tensors"
  input_side_packet: "CUSTOM_OP_RESOLVER:opresolver"
  options: {
    [mediapipe.TfLiteInferenceCalculatorOptions.ext] {
      model_path: "object_detection_3d.tflite"
    }
  }
}

# Decodes the model's output tensor (the heatmap and the distance fields) to 2D
# keypoints. There are nine 2D keypoints: one center keypoint and eight vertices
# for the 3D bounding box. The calculator parameters determine's the decoder's
# sensitivity.
node {
  calculator: "TfLiteTensorsToObjectsCalculator"
  input_stream: "TENSORS:detection_tensors"
  output_stream: "ANNOTATIONS:objects"
  options: {
    [mediapipe.TfLiteTensorsToObjectsCalculatorOptions.ext] {
      num_classes: 1
      num_keypoints: 9
      decoder_config {
        heatmap_threshold: 0.6
        local_max_distance: 2
        offset_scale_coef: 1.0
        voting_radius: 2
        voting_allowance: 1
        voting_threshold: 0.2
      }
      normalized_focal_x: 2.0975
      normalized_focal_y: 1.5731
    }
  }
}