HowtoCaffe>IR>TensorFlow>frozengraph?

Platform (like ubuntu 16.04/win10):
Mac OS X

Python version:
2.7

Source framework with version (like Tensorflow 1.4.1 with GPU):
Caffe

Destination framework with version (like CNTK 2.3 with GPU):
TensorFlow

Pre-trained model path (webpath or webdisk path):
Places365-CNN
Architecture, Weights

Running scripts:
I'm trying to convert the model to a frozen graph in TensorFlow but I'm missing a file. The steps I do are the following:

1. convert Caffe to IR

   1 2 3 4 5 6

   bash     python -m mmdnn.conversion._script.convertToIR \       -f caffe \       -n deploy_vgg16_places365.prototxt \       -w vgg16_places365.caffemodel \       -d ir_places365

   
   
   


2. convert IR to TF code

   1 2 3 4 5 6

   bash     python -m mmdnn.conversion._script.IRToCode \       -f tensorflow \       -n ir_places365.pb \       -w ir_places365.npy \       -d tf_places365.py

   
   
   


3. convert IR to TF model

   1 2 3 4 5

   bash     python -m mmdnn.conversion.examples.tensorflow.imagenet_test \       -n tf_places365.py \       -w ir_places365.npy \       --dump ./places365.ckpt

   
   
   


4. freeze graph

   1 2 3 4 5 6 7

   bash     python -m tensorflow.python.tools.freeze_graph \       --input_graph&＃061;places365.pb \ # What should I use as an input graph?       --input_binary&＃061;true \       --input_checkpoint&＃061;./places365.ckpt \       --output_graph&＃061;frozen_places365.pb \       --output_node_names&＃061;prob

   
   
   

该提问来源于开源项目：microsoft/MMdnn