Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

Model Downloader support for Tensor Flow

FARSHAD_A_Intel
Employee
724 Views

Hello,

I notice in model_downloader directory in deployment_tools, there are python scripts to convert Caffe2 and pytorch to ONNX. I wonder if there is any support for converting TensorFlow 1 and 2 to ONNX, as well? 

I also notice there are ONNX converters  at github.com/onnx. I wonder if these converters have been tested on Model Optimizer for compatibility?

thanks,

Farshad.

Labels (1)
0 Kudos
4 Replies
IntelSupport
Moderator
709 Views

Hi Farshad,


Thanks for reaching out.

TensorFlow Models conversion to IR is fully supported by OpenVINO Model Optimizer. We would encourage you to use the Model Optimizer conversion method.

The ONNX converter you mentioned may not work and you can give a try, but we cannot guarantee anything, as this converter has not been validated with Intel Distribution of OpenVINO Toolkit.

 

Please refer to the documentation in the following link:

https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow.html

 

Regards,

Aznie


0 Kudos
FARSHAD_A_Intel
Employee
687 Views

The response really does not answer my question. I do realize TF to IR is fully supported. The question I had was whether TF to ONNX is also supported by converter or model optimizer or other tools.

0 Kudos
IntelSupport
Moderator
667 Views

Hi Farshad,

 

The answer is no. OpenVINO Toolkit does not have any converter or Model Optimizer to convert the TensorFlow model into ONNX Model. But I have tried the converter from GitHub as you mentioned before and it is working. I successfully convert the TensorFlow model to the ONNX model.

 

Regards,

Aznie


0 Kudos
IntelSupport
Moderator
646 Views

Hi Farshad,


This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question. 


Regards,

Aznie


0 Kudos
Reply