I am new to openVino and want to integrate openvino with intel media sdk on linux.
is it possible to do so on Intel-VCA2 cards ?
Further more I want to create a pipeline like
Video_decode (on Media SDK Call , .i.e. MfxDecode Call) -> Some Computer Vision Analysis (OpenVino) -> Video_Encoder (on Media SDK Call , .i.e. MfxEncode Call)
At the same all these must be GPU accelerated as Intel Media SDK dose for decoding and Encoding. Is it possible?
Finally if you have any such application please share.
Hi Akash, K
Yes, we support VCA2 cards.
You can build E2E pipeline like you specified with MediaSDK and OpenVINO.
For now, you can't use GPU path for the whole pipeline.
There will be memcpy before the inference engine.
You can use MediaSDK for decoding, resizing, basic color format conversion, and encoding but you should do one more color format change from interleaved format to planar format for inference engine input.
We are providing MediaSDK stack in OpenVINO and will provide interop samples sooner or later.
Is there any sample application which has used msdk and openvino both. if so can you please provide details .
or can you guide me how can I create application which can use both ( msdk and openvino).
We're able to run Media SDK default sample executable's(without compilation).
We noticed below run-time errors when samples are compiled and run the above said samples.
We are using Openvino 2018.R5
We are able to see Vainfo as well
Can you pls look in to it and help us to resolve.