从导出的SavedModel形成图像有效载荷

Question

我已经使用TF对象检测API训练了模型。这将自动导出variables文件夹和saved_model.pb文件夹。我运行了saved_model_cli show --dir <saved_model_path> --all并得到了输出：

MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:

signature_def['serving_default']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['serialized_example'] tensor_info:
        dtype: DT_STRING
        shape: ()
        name: tf_example:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['detection_boxes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40, 4)
        name: detection_boxes:0
    outputs['detection_classes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40)
        name: detection_classes:0
    outputs['detection_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40)
        name: detection_scores:0
    outputs['num_detections'] tensor_info:
        dtype: DT_FLOAT
        shape: (1)
        name: num_detections:0
    outputs['raw_detection_boxes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 204800, 4)
        name: raw_detection_boxes:0
    outputs['raw_detection_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 204800, 2)
        name: raw_detection_scores:0
  Method name is: tensorflow/serving/predict

signature_def['tensorflow/serving/predict']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['serialized_example'] tensor_info:
        dtype: DT_STRING
        shape: ()
        name: tf_example:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['detection_boxes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40, 4)
        name: detection_boxes:0
    outputs['detection_classes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40)
        name: detection_classes:0
    outputs['detection_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 40)
        name: detection_scores:0
    outputs['num_detections'] tensor_info:
        dtype: DT_FLOAT
        shape: (1)
        name: num_detections:0
    outputs['raw_detection_boxes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 204800, 4)
        name: raw_detection_boxes:0
    outputs['raw_detection_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 204800, 2)
        name: raw_detection_scores:0
  Method name is: tensorflow/serving/predict

然后我用以下方法创建了一个预测函数：

predict_fn = tf.contrib.predictor.from_saved_model(<saved_model_path>)

现在，我被困在构造要发送给它的图像有效载荷上。特别是，我对serialized_example的输入名称，DT_STRING的输入dtype和()的输入形状感到困惑。有人愿意演示如何获取numpy数组或jpeg图像文件并为该模型形成适当的有效载荷吗？