CPU plugin error person-vehicle-bike-detection-crossroad-0078 on Raspian

[Robinson__Brian]

I have installed Openvino on my Raspberry pi 3 B+ (with Raspian Stretch) according to the instructions and have successfully run both test demos. I built the intel model person-vehicle-bike-detection-crossroad-0078 and pulled the pretrained network and weights from the download site. I am calling the inference engine with the following:

./crossroad_camera_demo -d MYRIAD -i ./ped_veh_test.jpg -m person-vehicle-bike-detection-crossroad-0078.xml

and I get the following output and error:

InferenceEngine: 
    API version ............ 1.4
    Build .................. 19154
[ INFO ] Parsing input parameters
[ INFO ] Reading input
[ INFO ] Loading plugin MYRIAD

    API version ............ 1.5
    Build .................. 19154
    Description ....... myriadPlugin
[ INFO ] Loading plugin CPU

[ ERROR ] Cannot find plugin to use :
 

So it apparently finds the MYRIAD plugin. I also tried including paths to the plugin libraries (I tried all libraries in the directory, which includes the libmyriadPlugin.so), but it gave the same error:

./crossroad_camera_demo -d MYRIAD -i ./ped_veh_test.jpg -m person-vehicle-bike-detection-crossroad-0078.xml -l /home/pi/Downloads/inference_engine_vpu_arm/deployment_tools/inference_engine/lib/raspbian_9/armv7l/libcpu_extension.so

Am I missing the MKLDNN plugin for armv7 or something?

Thank you.

[Shubha_R_Intel]

Dearest Brian:

If this is the exact command you typed:

./crossroad_camera_demo -d MYRIAD -i ./ped_veh_test.jpg -m person-vehicle-bike-detection-crossroad-0078.xml -l /home/pi/Downloads/inference_engine_vpu_arm/deployment_tools/inference_engine/lib/raspbian_9/armv7l/libcpu_extension.so

libcpu_extension.so should be lib/cpu_extension.so (note the forward slash).

Thanks for using OpenVino !

Shubha

[Robinson__Brian]

Dearest Shubha,

Thank you for taking the time to answer but that is not a valid path. Please look again. On the other hand, I've never had to supply a path to a shared object file. That has always been the job of the linker. So I'm probably misinterpreting the help output. If someone could take a look I would appreciate it.

Thanks,

Brian

[Shubha_R_Intel]

Dearest Brian, sorry for misunderstanding and I definitely see your point. 

Let me try and get an appropriate answer for you.

Thanks for your patience and also for using OpenVIno !

Shubha

[Shubha_R_Intel]

Dearest Brian,

I have filed a bug on your behalf. 

It should not be looking for a CPU Plugin when you specified -d MYRIAD. I'm sorry for the inconvenience this has caused but congratulations - you found an OpenVino bug ! It will be fixed for the next release.

Thanks for using OpenVino,

Shubha

[Wolter__Andreas]

Hello,

I tried the same demo and got the same error.

I am using a Raspberry Pi 3B+ with NCS2.

I have installed l_openvino_toolkit_raspbi_p_2019.1.094.tgz

I think that didnt was fixed in this version. Could you tell us a workaround?

I am searching for an example for using live camera support. The default example object_detection_sample_ssd is using still images. I spend much time to bring the yoloV3 example to work. It works now and I want to compare the realtime person detection with other algorithms.

Thanks.

Reagards,
Andreas

[Wolter__Andreas]

Hello,

the same error occures with interactive face detection demo if a second model was set with -m_ag. if only the -m option is set the demo works and detect faces. I think there is an error in the command line parser.

Why is there a for loop to iterate over the plugins? Is this an option if we want the MYRIAD and CPU work at the same time?

Regards,

Andreas

[Shubha_R_Intel]

Dear Wolter, Andreas,

for live camera support samples, look at our samples with "_async" in the title:

In C++, you will find:

classification_sample_async, object_detection_demo_yolov3_async

In Python you will find:

classification_sample_async, object_detection_demo_ssd_async, object_detection_demo_yolov3_async

Hope it helps.

Thanks,

Shubha