OpenCV optimization on raspberry pi?

[Matsuzawa__Takashi]

I am looking into any optimization that can be applied to OpnVINO + RPi easily, without much change in the code.

Google'ing for hints, I can see people first tries the following with OoenCV setup.

1) Enable ARM compiler flags (NEON, VFP)
2) Enable TBB

Raspberry Pi version of OpenVINO kit is provided with binary libraries.
I wonder how they are specified with these options?

And, if TBB not yet enabled, is it a good idea that I install TBB package onto my RPi?
(Though not sure OpenVINO binary libraries change their behavior after that, without rebuild from the source.)
 

[Maksim_S_Intel]

You can check output of cv::getBuildInformation() function (cv2.getBuildInformation() in Python). It says: NEON enabled, parallel framework - pthreads.

  CPU/HW features:
    Baseline:                    NEON
      required:                  NEON
      disabled:                  VFPV3

If you want to use OpenCV with TBB parallel framework, you need to build it from sources.

[Matsuzawa__Takashi]

Thanks, I think this is enough for me at this time.