tebesu / tensorflow-cpp-api-training Goto Github PK

Can I please know how to save graph in c++ after the training is completed

Hello, thank u for ur tensorflow C++ code!
How can I save the model after training?
Is there a method like saver in python API to save the model into a ckpt file or .pb file?

Hi
I am doing research on face Recognition using tensorflow
Can I please know how how to use the code to create model file for face recognition.

Task (required in C++)
I have 5 folders, each folder has 4 images of a particular person.
I would like to train the images and generate a model file from scratch.
I would then like to use the model file for recognizing one of the 5 persons

Steps followed

1. https://www.tensorflow.org/tutorials/image_retraining
   Issues

2. Its in Python , but I wanted in C++

3. It is not training the model file from scratch

4. https://github.com/tensorflow/models/tree/master/research/inception/inception
   Issue Resolved - Model can be trained from scratch
   Unresolved Issue - Its in Python , but I wanted in C++

Please help me in creating the model file in C++

i am testing your code on Xcode 8.2 with mac 10.11.6.
TF_CHECK_OK(ReadBinaryProto(Env::Default(), graph_definition, &graph_def)); this line gives me a crash.
Check failed: ::tensorflow::Status::OK() == (ReadBinaryProto(Env::Default(), graph_definition, &graph_def)) (OK vs. Not found: mlp.pb)

Any idea why it is doing so?

First, thank you for the post I'm trying similar things with the C++ API. However, when reading through the blog post it wasn't clear to me what the model actually does. Is it just adding 32 random numbers generated between 0 to 1 together?