topic SUCCESS with a car and plate in Intel® Distribution of OpenVINO™ Toolkit

Open Vino Demo fails building

RTasa — Fri, 07 Dec 2018 16:26:30 GMT

Build Inference engine samples using Visual Studio (MSBuild.exe) error C3861: ...softmax.h _mm_blendv.ps : identifier not found [... ie_cpu_extension.vcxproj]

What version of Visual Studio

nikos1 — Fri, 07 Dec 2018 19:32:50 GMT

What version of Visual Studio are you using?

Any issues from cmake?

Version is 2017. Building

RTasa — Fri, 07 Dec 2018 21:36:00 GMT

VS 2017 Windows 10 . Building before that seemed smooth. My Laptop for this is not the latest would that have anything to do with the build of the demo?

Yes could be related to an

nikos1 — Fri, 07 Dec 2018 21:52:00 GMT

Yes could be related to an old CPU that does not support some SSE instruction.

What is the CPU model? Could you try on a more recent system?

Just found this that looks relevant (?)

https://medium.com/developerofstuff/openvino-mm-blendv-ps-build-error-bdc37855ac04

Maybe try to edit C:\Intel\computer_vision_sdk_2018.4.420\deployment_tools\inference_engine\src\extension\common\defs.h

#if defined (HAVE_SSE) || defined (HAVE_AVX2)
#if defined (_WIN32)
// #include <emmintrin.h>
#include <smmintrin.h>
#else
#include <x86intrin.h>
#endif
#endif

I just downloading and

RTasa — Fri, 07 Dec 2018 22:27:20 GMT

I just downloading and installing CPU-Z and found out I have an i 5 Ivy Bridge which supports SSE 4.2 So I am going to change the DEF file and see what happens.

SUCCESS with a car and plate

RTasa — Fri, 07 Dec 2018 22:34:50 GMT

SUCCESS with a car and plate detected. Unless it was faked this is cool. Thanks so so much. I tried searching for the error cut did not find the blog post.

Good to know! Not faked.. :-

nikos1 — Fri, 07 Dec 2018 22:51:53 GMT

Good to know! Not faked.. :-) I tried on other clips too. The pre-trained OpenVino models are pretty accurate for Chinese cars/plates.

Cheers,

Nikos

The other sample made me feel

RTasa — Fri, 07 Dec 2018 23:08:54 GMT

The other sample made me feel OK I got a 38.x fps inference which is very decent. Feels like I have so much to learn I did some Tensorflow before now trying this with Movidius Stick version 2