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

Can the Model Optimizer convert this Keras Transfer Traning models to IR

pang__pih
Beginner
258 Views

The Keras model is pretrained with

vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5 and

inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5.

Is there anyway I can convert this trained model to IR, and use a NCS2 as Inference Engine?

Does the Model Optimizer support all kinds TensorFlow2 trained models, or is it only YOLO/SSD network strutures,

and does it accept weights in .h5 format?

Hope you can help me answer these questions, Thanks?

Pih Lung

 

 

More info regarding the Keras model see attached files and the url

https://www.kaggle.com/paultimothymooney/interpret-sign-language-with-deep-learning

 

0 Kudos
1 Reply
JAIVIN_J_Intel
Employee
258 Views

Hi Pih,

The Model Optimizer expects the models to be one of the following supported frameworks:

TensorFlow, Caffe, Kaldi, ONNX or MXnet

Please convert the Keras model to TensorFlow and then use the Model Optimizer.

You may refer Supported Topologies to see the list of topologies supported with TensorFlow* framework.

Also, verify that the model contains the TensorFlow* Supported Operations before converting to IR.

Regards,

Jaivin

Reply