Intel® Compute Stick
Support for Intel® Compute Sticks and Cards
531 Discussions

i hvae run demo on raspberry with the same model but different result as below. why

BBi
Beginner
632 Views

when i followed the installation of openvino on raspberry official site to install openvino and download the demo model face-detection-adas-0001.xml. then i use ncs2 with the demo object_detection_sample_ssd but failed at last as below info:

 pi@raspberrypi:~ild $ ./armv7l/Release/object_detection_sample_ssd -m ~/Downloads/face-detection-adas-0001.xml -d MYRIAD -i ~/Downloads/pic/car.png       [ INFO ] InferenceEngine:

        API version ............ 2.1

        Build .................. custom_releases/2020/1_d349c3ba4a2508be72f413fa4dee92cc0e4bc0e1

        Description ....... API

Parsing input parameters

[ INFO ] Files were added: 1

[ INFO ]     /home/pi/Downloads/pic/car.png

[ INFO ] Loading Inference Engine

[ INFO ] Device info:

        MYRIAD

        myriadPlugin version ......... 2.1

        Build ........... 37988

[ INFO ] Loading network files:

        /home/pi/Downloads/face-detection-adas-0001.xml

        /home/pi/Downloads/face-detection-adas-0001.bin

[ ERROR ] Check 'axis < static_cast<size_t>(input_rank)' failed at /teamcity/work/scoring_engine_build/releases_2020_1raphcraph/op/gather.cpp:140:

While validating node 'Gather[Gather_852](patternLabel_848: float{10,20,30}, patternLabel_849: int64_t{5}, patternLabel_851: int64_t{1}) -> (??)':

The axis must => 0 and <= input_rank (axis: 4294967295).

 

then i replace the model with the same model i downloaded in 2019, run it and it's ok as below info:

 [90,1] element, prob = 0.0273438    (-4,47)-(18,82) batch id : 0

[91,1] element, prob = 0.0273438    (7,119)-(40,151) batch id : 0

[92,1] element, prob = 0.0273438    (82,143)-(120,187) batch id : 0

[93,1] element, prob = 0.0273438    (60,156)-(100,198) batch id : 0

[94,1] element, prob = 0.0273438    (42,1)-(137,72) batch id : 0

[95,1] element, prob = 0.0273438    (656,67)-(725,123) batch id : 0

[96,1] element, prob = 0.0273438    (48,118)-(127,163) batch id : 0

[ INFO ] Image out_0.bmp created!

[ INFO ] Execution successful

 

[ INFO ] This sample is an API example, for any performance measurements please use the dedicated benchmark_app tool

pi@raspberrypi:~ild $ ./armv7l/Release/object_detection_sample_ssd -m ~/Downloads/2019model16/face-detection-adas-0001.xml -d MYRIAD -i ~/Downloads/pic/car.png 

 

so, i don't know the reason. Have you ever encountered this problem and found the detailed reason?

 

0 Kudos
4 Replies
BBi
Beginner
423 Views

add more info:

the difference between the two---

root@raspberrypi:/home/pi/Downloads# md5sum face-detection-adas-0001.xml

05ad37b2fbc9c6c813387aa0e9745a6d face-detection-adas-0001.xml

root@raspberrypi:/home/pi/Downloads# md5sum 2019model16/face-detection-adas-0001.xml

ce5190baff975294a34c36414196cfc5 2019model16/face-detection-adas-0001.xml

root@raspberrypi:/home/pi/Downloads#

 

JAVIERJOSE_A_Intel
423 Views

Hi BBi,

 

Thanks for reaching out.

 

This is an issue with the latest OpenVINO™ toolkit version (2020.1). The "axis" error is caused by the new IR v10 format. 

 

As a workaround for this you can:

 

  1. Use the previous pre-trained models from the 2019 R3 open model zoo.
  2. Generate the IRv7 format files using the Model Optimizer from the full installation of OpenVINO™ toolkit with the following parameter: --generate_deprecated_IR_V7
  3. Install the previous version of OpenVINO™ toolkit (2019 R3.1) on a Windows, Linux or Mac OS system, convert the model there, which will generate the IR v7 format and that should work successfully with the Intel® Neural Compute Stick 2 on the Raspberry Pi.

 

 

Regards,

 

Javier A.

BBi
Beginner
423 Views
Hi, Javier, Got it Thank you very much for your kindly reply
JAVIERJOSE_A_Intel
423 Views

Hi BBi,

 

We are happy we could help you. If you have additional questions, let us know.

 

Regards,

 

Javier A.

Reply