Inference on openvino for image enhancement workload

[Aswathy_C_Intel]

Hi Team,

I am trying to do the inference for an image enhancement workload in openvino. I could convert the model into IR. I tried to do the inference for the same using the ‘style_transfer_sample’  in python samples of openvino inference engine. But I am facing multiple issues.

Error:

assert len(net.outputs) == 1, "Sample supports only single output topologies"
AssertionError: Sample supports only single output topologies

I commented the assertions and tried running the same and now faces the below error.

"Invalid number of channels in input image:
>     'VScn::contains(scn)'"

Could you please let me know whether this sample (style_transfer_sample) is recommended for such a workload? If so, what all are the steps/changes I should make to do the inference without errors.

Thanks in advance.

[HemanthKum_G_Intel]

Hi Aswathy,

The OpenVINO™ toolkit does not include a pre-trained model to run the Neural Style Transfer sample. A public model from the Zhaw's Neural Style Transfer repository can be used.  To use the style transfer sample from OpenVINO™, follow the steps below as no public pre-trained style transfer model is provided with the OpenVINO toolkit:- https://docs.openvinotoolkit.org/2019_R3/_docs_MO_DG_prepare_model_convert_model_mxnet_specific_Convert_Style_Transfer_From_MXNet.html

The style transfer sample supports only single output topologies, which means the model should have single output layer as demonstrated in the above link.