无法将冻结的模型文件 .pb 转换为 .tflite

Question

import tensorflow as tf

path = "C:/Users/LAWSSSS/Desktop/convert_pb_2_tflite/frozen_inference_graph-SteelRoll.pb"

inputs = ["image_tensor"]
outputs = ["detection_boxes"]

converter = tf.lite.TFLiteConverter.from_frozen_graph(path, inputs, outputs, input_shapes={"image_tensor":[1,640,360,3]})
converter.post_training_quantize = True
tflite_model = converter.convert()

open("frozen_inference_graph-SteelRoll.tflite", "wb").write(tflite_model)

总的来说，我是 Tensorflow 的新手。我尝试使用上面的代码将 .pb 转换为 .tflite。但是，我收到一个错误：

array.data_type == array.final_data_type Array "image_tensor" has mis-matching actual and final data types (data_type=uint8, final_data_type=float).
Fatal Python error: Aborted

我该如何解决这个问题？

Answer 1

将模型输入（im​​age_tensor 占位符）更改为数据类型为 tf.float32。