TensorFlow Lite + PC

环境

• Win10 x86 64
• TensorFlow 2.3.1
• Python3.6

目录

安装tflite

pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp36-cp36m-win_amd64.whl

tflite下载地址平台版本对应

本机是intel的cpu但安装amd64也可以正常运行，毕竟windows里就没有x86的选项。

模型转换

将基于 TensorFlow 2.3.1训练得到的模型直接转换为 TensorFlow Lite 模型，官方指南，我的TensorFlow 模型是PB格式，转换代码如下：

import tensorflow as tf

if __name__ == '__main__':
    int_path = "TensorFlow_model_path_pb"
    out_path = ""

    # Converting a SavedModel to a TensorFlow Lite model.
    converter = tf.lite.TFLiteConverter.from_saved_model(int_path )
    #Normalization or quantization
    converter.post_training_quantize = True
    tflite_model = converter.convert()
    print(tflite_model)
    open(out_path, "wb").write(tflite_model)

其中有一个选项是Normalization and quantization pa