Keras模型转成tensorflow的.pb操作

Keras的.h5模型转成tensorflow的.pb格式模型，方便后期的前端部署。直接上代码

from keras.models import Model
from keras.layers import Dense, Dropout
from keras.applications.mobilenet import MobileNet
from keras.applications.mobilenet import preprocess_input
from keras.preprocessing.image import load_img, img_to_array
import tensorflow as tf
from keras import backend as K
import os

base_model = MobileNet((None, None, 3), alpha=1, include_top=False, pooling='avg', weights=None)
x = Dropout(0.75)(base_model.output)
x = Dense(10, activation='softmax')(x)

model = Model(base_model.input, x)
model.load_weights('mobilenet_weights.h5')

def freeze_session(session, keep_var_names=None, output_names=None, clear_devices=True):
 from tensorflow.python.framework.graph_util import convert_variables_to_constants
 graph = session.graph
 with graph.as_default():
  freeze_var_names = list(set(v.op.name for v in tf.global_variables()).difference(keep_var_names or []))
  output_names = output_names or []
  output_names += [v.op.name for v in tf.global_variables()]
  input_graph_def = graph.as_graph_def()
  if clear_devices:
   for node in input_graph_def.node:
    node.device = ""
  frozen_graph = convert_variables_to_constants(session, input_graph_def,
             output_names, freeze_var_names)
  return frozen_graph

output_graph_name = 'NIMA.pb'
output_fld = ''
#K.set_learning_phase(0)

print('input is :', model.input.name)
print ('output is:', model.output.name)

sess = K.get_session()
frozen_graph = freeze_session(K.get_session(), output_names=[model.output.op.name])

from tensorflow.python.framework import graph_io
graph_io.write_graph(frozen_graph, output_fld, output_graph_name, as_text=False)
print('saved the constant graph (ready for inference) at: ', os.path.join(output_fld, output_graph_name))

补充知识：keras h5 model 转换为tflite

在移动端的模型,若选择tensorflow或者keras最基本的就是生成tflite文件，以本文记录一次转换过程。

环境

tensorflow 1.12.0

python 3.6.5

h5 model saved by `model.save('tf.h5')`

直接转换

`tflite_convert --output_file=tf.tflite --keras_model_file=tf.h5`
output
`TypeError: __init__() missing 2 required positional arguments: 'filters' and 'kernel_size'`

先转成pb再转tflite

```

git clone git@github.com:amir-abdi/keras_to_tensorflow.git
cd keras_to_tensorflow
python keras_to_tensorflow.py --input_model=path/to/tf.h5 --output_model=path/to/tf.pb
tflite_convert \

 --output_file=tf.tflite \
 --graph_def_file=tf.pb \
 --input_arrays=convolution2d_1_input \
 --output_arrays=dense_3/BiasAdd \
 --input_shape=1,3,448,448
```

参数说明，input_arrays和output_arrays是model的起始输入变量名和结束变量名，input_shape是和input_arrays对应

官网是说需要用到tenorboard来查看，一个比较trick的方法

先执行上面的命令，会报convolution2d_1_input找不到，在堆栈里面有convert_saved_model.py文件，get_tensors_from_tensor_names()这个方法，添加`print(list(tensor_name_to_tensor))` 到 tensor_name_to_tensor 这个变量下面，再执行一遍，会打印出所有tensor的名字，再根据自己的模型很容易就能判断出实际的name。

以上这篇Keras模型转成tensorflow的.pb操作就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持我们。