Converting OmniScaleNet from PyTorch to Intel IR

[Socke__Linux]

Is there already something in the model optimizer to convert a self learned OSNet?

[SuryaPSC_Intel · Accepted solution]

Hi Linux,

OSNet is not an OpenVINO pytorch supported topology. Kindly refer to Supported Pytorch* Models via ONNX Conversion.

Best Regards,

Surya

[Socke__Linux]

Ok, thanks.

Can give me some links where I can learn to do it myself then?

Greetings Socke

[Max_L_Intel]

Hi Socke, Linux

As Surya mentioned above, OSNet is not the one of officially supported topologies. However, you can try to convert this via ONNX conversion if all the model layers are supported by Model Optimizer, as per https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Supported_Frameworks_Layers.html#onnx_supported_operators
and your inference device, as per  https://docs.openvinotoolkit.org/latest/_docs_IE_DG_supported_plugins_Supported_Devices.html#supported_layers

The command is:

python3 mo.py --input_model <INPUT_MODEL>.onnx

Thanks.
Best regards, Max.