mreliptik / keras_to_tf_ncs2

Keras to Tensorflow test for Neural Compute Stick 2

Keras to Tensorflow to IR model (NCS2)

Goal

Convert a Keras or a Tensorflow model to IR files ready to be used with the Neural Compute Stick 2

Prerequisites

For that you want to have OpenVino installed and python 3.5 at least. For the python requirements, see the "Requirements" section.

File structure

How to use

Keras to Tensorflow conversion

If you have a Keras .h5 file, use keras_to_tf.py to create a Tensorflow .pb file.

    python keras_to_tf.py

that will take the Keras file situated in Keras_model/model.h5 and create a .pb file in TF_model/tf_model.pb.

Tensorflow to IR conversion

If you didn't had a .pb model before now you should have one. We'll use the model optimizer to convert the file.

    mo.py --data_type FP16 --framework tf --input_model TF_model/tf_model.pb --model_name IR_model --output_dir IR_model/ --input_shape [1,28,28,1] --input conv2d_1_input --output activation_6/Softmax

Runing the inference on the NCS2

Now you can run the inference on the NCS2. For that use the predict_mnist.py

    python predict_mnist.py

This file load the IR model, read and convert the data/6.jpg and feed it for classification.

If everything goes fine, you should see something like this:

    [ INFO ] Loading network files:
    IR_model/IR_model.xml
    IR_model/IR_model.bin
    [ INFO ] Preparing input blobs
    1 1 28 28
    (28, 28)
    (1, 28, 28)
    [ INFO ] Loading model to the plugin
    [ INFO ] Starting inference (1 iterations)
    [ INFO ] Average running time of one iteration: 1.8284320831298828 ms
    [ INFO ] Processing output blob
    [[0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]]

The last line is the class vector. We have a 1 at index 6, so the image has been correctly classified.

Requirements

    pip install -r requirements.txt

keras_to_tf_ncs2's Issues

TypeError: Keras symbolic inputs/outputs do not implement `op`. You may be trying to pass Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing Keras from automatically converting the API call to a lambda layer in the Functional Model.

Warning (from warnings module):
File "C:\Users\Admin\AppData\Local\Programs\Python\Python39\lib\site-packages\keras\backend.py", line 414
warnings.warn('tf.keras.backend.set_learning_phase is deprecated and '
UserWarning: tf.keras.backend.set_learning_phase is deprecated and will be removed after 2020-10-11. To update it, simply pass a True/False value to the training argument of the __call__ method of your layer or model.
WARNING:tensorflow:No training configuration found in the save file, so the model was not compiled. Compile it manually.
Traceback (most recent call last):
File "E:\PYTHON\KERAS TO TF NCS\Keras_to_TF_NCS2-master\keras_to_tf.py", line 43, in
output_names=[out.op.name for out in model.outputs])
File "E:\PYTHON\KERAS TO TF NCS\Keras_to_TF_NCS2-master\keras_to_tf.py", line 43, in
output_names=[out.op.name for out in model.outputs])
File "C:\Users\Admin\AppData\Local\Programs\Python\Python39\lib\site-packages\keras\engine\keras_tensor.py", line 233, in op
raise TypeError('Keras symbolic inputs/outputs do not '
TypeError: Keras symbolic inputs/outputs do not implement op. You may be trying to pass Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing Keras from automatically converting the API call to a lambda layer in the Functional Model.