https://github.com/mreliptik/keras_to_tf_ncs2

Keras to Tensorflow test for Neural Compute Stick 2
https://github.com/mreliptik/keras_to_tf_ncs2

keras model-optimizer ncs2 openvino tensorflow

Last synced: 10 days ago
JSON representation

Keras to Tensorflow test for Neural Compute Stick 2

• Host: GitHub
• URL: https://github.com/mreliptik/keras_to_tf_ncs2
• Owner: MrEliptik
• Created: 2019-03-02T13:44:16.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2022-11-22T01:06:43.000Z (almost 2 years ago)
• Last Synced: 2023-03-03T00:08:18.075Z (over 1 year ago)
• Topics: keras, model-optimizer, ncs2, openvino, tensorflow
• Language: Python
• Size: 5.46 MB
• Stars: 9
• Watchers: 1
• Forks: 5
• Open Issues: 15
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

README

        
# 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