Multiple model on NCS2 in Async Fashion

[SSola8]

I have gone through many samples code and read about all Openvino Python API. I am concern about how can I load multiple model in async fashion on One NCS2. I know that from here, that Each IExecutableNetwork instance tries to allocate new device on InferenceEngine::Core::LoadNetwork, but How can we allocate multiple models on one NCS2 in async fashion(not pipelined ) same as we do in one model on one NCS2 by changing num_requests parameter. Do we have anything such here? Is it possible in NCS1 or NCS2? 

In short, Is Multiple model on NCS2 in Async fashion or parallel processing possible?

[SuryaPSC_Intel]

Hi Solanki, Saurav,

Yes, you can deploy multiple models in Async fashion on NCS2.

It can be done similarly as we load a single model.

You can set number of infer requests for each model separately using num_requests parameter.

Kindly refer security_barrier_camera_demo which supports three models in async mode and can also be deployed on a single NCS2.

Best Regards,

Surya

[SSola8]

Hi @Chauhan, Surya Pratap Singh,

Can you give a related example to python as well?

[SuryaPSC_Intel]

Hi Solanki, Saurav,

You can refer to action recognition demo (python) which supports two models in async mode and can also be deployed on a single NCS2.

Best Regards,

Surya