Neural compute stick 2 for raspberry pi 3 B+?

[khandelwal__Yash]

I purchased  NCS2 for running it pi but after purchasing only i get to know it does'nt support raspbian.

Will intel make it for pi also in future?

please tell i am a student and i payed my whole savings in buying NCS2.

[fu__xiaocong]

Nunez-Yanez, Jose wrote:

I followed your instructions but my system is ubuntu 16.04 on a 64bit ARM board. 

When I do the final build_samples.sh it errors out trying to link:

root/openvino/inference_engine_vpu_arm/deployment_tools/inference_engine/lib/ubuntu_16.04/aarch64/libinference_engine.so: error adding symbols: File in wrong format
 

Seems obvious that the .so is 32bit but the generated code is 64bit so the cannot be linked.  It seems that this is not a problem in your debian system ? 

Hi,

I had the same error when I used Debian arm64 rootfs, so I switched to armhf. I think it's no complicated to change the rootfs. You can refer to my steps:

$ sudo tar -xzvf linaro-stretch-developer-20170706-43.tar.gz

Insert sd card to linux machine and check with ls /dev/sd*

$ sudo umount /dev/sdx

$ sudo mkfs.ext4 /dev/sdx

$ mkdir rootfs

$ sudo mount /dev/sdx rootfs

$ sudo cp -a -r binary/* rootfs

$ sudo umount /dev/sdx

[Lee__Edward]

I have followed this guide to install OpenVINO on Raspberry Pi 3.
When I ran the sample object_detection_sample_ssd with model face-detection-adas-0001, I found that the performance of Myriad X is just about 7 FPS.
I also tried to run the same sample by using Myriad 2, and the performance is about 6 FPS.

Then I tried to run the sample benchmark_app with model face-detection-adas-0001, and the performance of Myriad X is about 9 FPS.
I though the performance of Myriad X should be much better than Myriad 2.
Did I do something wrong?

[Marcia_W_Intel]

For the bitness message - change the path in setupvars.sh to include the directory /inference_engine_vpu_arm

I can confirm I did receive this message regarding the rules at one point but do not recall my fix. Most likely a path correction. 

Jelbert, Richard wrote:

I have also just purchased a NCS2 and found it is not compatible with the Raspberry Pi. I followed the instructions in the post above by downloading and modifying various files and environment variables (from here: https://software.intel.com/en-us/articles/OpenVINO-Install-RaspberryPI )

I am on a fresh Rpi install reporting:Linux turbopi 4.14.79-v7+ #1159 SMP Sun Nov 4 17:50:20 GMT 2018 armv7l which is raspbian stretch

The first problem I spotted when following the install instructions is that there are two copies of setupvars.sh :

/home/pi/Downloads/inference_engine_vpu_arm/bin/setupvars.sh
/home/pi/Downloads/inference_engine_vpu_arm/opencv/setupvars.sh

... so I followed the instructions and only updated the one in /bin and not in opencv. I changed the path to /home/pi/ncs2/

after modifying the .bashrc I now get an error message:

Last login: Thu Jan  3 10:12:30 2019 from 192.168.0.27
[setupvars.sh] 64 bitness for Python 3.5 is requred
[setupvars.sh] OpenVINO environment initialized

... I can not find any reference to the 64 bitness error but the next message says OpenVINO is initialized so I continue...

when I try and run the rules script :

sh ~/Downloads/inference_engine_vpu_arm/install_dependencies/install_NCS_udev_rules.sh

... I get an error saying:

Update udev rules so that the toolkit can communicate with your neural compute stick
File '97-myriad-usbboot.rules' is missing. Please check that you installed 'Inference Engine Runtime for Intel® Movidius™ VPU'.
/home/pi/Downloads/inference_engine_vpu_arm/install_dependencies/install_NCS_udev_rules.sh: 30: exit: Illegal number: -1

... so unfortunately I can't continue to test the examples.

Are you able to identify the errors I am seeing?

Would it be possible to write a bash script to automatically modify the toolkit and update the things necessary?

I'd like to set up a cluster of RPi's with Neural Compute Stick v2's

Many thanks for any help

Richard