0

I want to export (optimize) a TensorFlow 2 model to OpenVino.

The only documentation I found regards Tensorflow 1. When followed the instructions, the OpenVino model optimization failed to work with a tf2 model.

Gal Hyams
  • 336
  • 1
  • 5
  • 15
  • TensorFlow 2.x is not currently supported by the model optimizer (see [this forum thread](https://software.intel.com/en-us/forums/computer-vision/topic/827537)). According to the [Model Optimizer Developer Guide](https://docs.openvinotoolkit.org/latest/_docs_MO_DG_Deep_Learning_Model_Optimizer_DevGuide.html), the latest release is compatible with 1.14.0. – jdehesa Nov 11 '19 at 15:29

1 Answers1

2

OpenVino Model Optimizer does not support Tensorflow 2.0 yet. But, you can use Tensorflow 1.14 freeze_graph.py to freeze a TF 2.0 model. This frozen pb should be accepted by Model Optimizer.

  • Just trying to confirm here. With `tf.keras` in TensorFlow 2.0, when we do `model.save("your_directory"), it does generate a `.pb` file. Is it accepted by OpenVINO? – S. P Feb 23 '20 at 14:10