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.
6392 Discussions

Export Keras VGG16 fine tuned model to Movidius

idata
Employee
815 Views

Hello,

 

I have fine tuned the Keras implementation of the VGG16 net loaded with the following code:

 

vgg16 = VGG16(weights="imagenet", include_top=False, input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3))

 

After replacing the classification layers and performed the training, I have exported the json model and the weights to convert the net for tensorflow.

 

Then, I have exported the checkpoint, .data, .index and .meta files using the tensorflow functions.

 

I have all the file necessary to create the model compatible with Movidius with the command:

 

mvNCCompile tf_export/garbage_vgg16.meta -in=vgg16_input -on=dense_2/Softmax

 

But the command fails with the following error:

 

[Error 5] Toolkit Error: Stage Details Not Supported: Switch

 

How can I fix this problem?

 

Thanks in advance for your help.

0 Kudos
1 Reply
idata
Employee
495 Views

Hi @macc.n

 

Can you please post the whole output?

 

"Stage details not supported" errors usually mean that your network requires operations or layers that aren't yet supported by the NCSDK. If you enable debugging (change to debug = true in /ncsdk/Controllers/TensorFlowParser.py) you can see exactly which layers or operations are unsupported.

 

Best Regards,

 

Sahira
0 Kudos
Reply