Community
cancel
Showing results for 
Search instead for 
Did you mean: 
li__lang
Beginner
120 Views

OpenVINO 2018.R3 Model Optimizer raises error when processing Tensorflow Stridedslice Op

1. I am using the model optimizer to convert a custom tensorflow model that has a strided slice node. Please see the attachments for the debug output. How to solve this problem?  Is waiting for the update of OpenVINO 2018.R4?

2. And the bottom of the website (https://software.intel.com/en-us/articles/OpenVINO-Using-TensorFlow) says that "Some TensorFlow* operations don't match to an Inference Engine layer, but are still supported by the Model Optimizer and can be used on constant propagation path. " How to use the oprations on "constant propagation path"?

0 Kudos
3 Replies
Monique_J_Intel
Employee
120 Views

Hi Lang,

The error that Model Optimizer(MO) is outputting is due to MO not fully interpreting the custom Strided slice layer that you have in the model. You will need to create a custom layer for the Strided slice layer in your model. Please follow in package documentation(/opt/intel/computer_vision_sdk/deployment_tools/documentation/docs/CustomLayersExtension.html) to do so.

In the meantime we are working on the TensorFlow strided slice layer and I will check to see if the implementation will be ready by R4.

Kind Regards,

Monique Jones

Monique_J_Intel
Employee
120 Views

Hi Lang,

Since there are many varieties of strided slice node I will need to look at your model further to understand if the case that Model Optimizer isn't handling is one of the cases that will be handled in the R4 release. Could you please attach your model?

Kind Regards,

Monique Jones

li__lang
Beginner
120 Views

Hi Monique,

I'm sorry to reply you so late. My model's link is https://pan.baidu.com/s/1RqUn0NBGd7JcF_vRJTMmSg, and my extraction code is mu1y. (open the website and input the code)

If you cannot download it, please contact me!

My script:

sudo python mo_tf.py --input_model model_iv4.pb \
--input Placeholder_X \
--input_shape [1,299,299,3] \
--mean_values [127.5,127.5,127.5] \
--output Predictions 
 
Regards,
Monique 
Reply