Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
158 Views

I found out that the intel NCS2 plug in USB3 slot on raspberry pi 4 and then run the object-detection program, the result will be freezed and present 'Network error: Software caused connection abort'. I need to turn off power and restart pi4.

However, while the NCS2 plug in USB2 slot on pi 4, it works well.

 

Can you fix this? Thanks you so much.

0 Kudos
8 Replies
Highlighted
Employee
59 Views

Hi User15891561295748856269,

 

Thanks for reaching out. We tested an NCS2 on a Raspberry Pi 4 Model B, connecting it to the USB 3.0  port and it worked successfully. Could you try installing a fresh version of Raspbian OS and see if the error persists?

 

Regards,

David C.

0 Kudos
Highlighted
59 Views

ok, I had reinstalled a fresh version of Raspbian OS last night and executed the same program.

 

On USB 2 is perfect, no any error. However, the freezed still happened on USB 3 slot. I execute the Ctrl + C, the error message show below.

 

E: [global] [   829812] [Scheduler00Thr] dispatcherEventSend:1013     Write failed -1

 

E: [global] [   829975] [Scheduler00Thr] dispatcherEventSend:1004     Write failed (header) (err -1) | event XLINK_WRITE_REQ

 

E: [xLink] [   829975] [Scheduler00Thr] eventSchedulerRun:626 Event sending failed

E: [xLink] [   841783] [EventRead00Thr] eventReader:218       eventReader thread stopped (err -4)

E: [xLink] [   841783] [Scheduler00Thr] eventSchedulerRun:576 Dispatcher received NULL event!

E: [ncAPI] [   841784] [python] ncGraphQueueInference:3973    Can't send trigger request

E: [watchdog] [   841784] [WatchdogThread] sendPingMessage:121 Failed send ping message: X_LINK_ERROR

terminate called after throwing an instance of 'InferenceEngine::details::InferenceEngineException'

 what(): Failed to queue inference: NC_ERROR

 

0 Kudos
Highlighted
59 Views

I changed to use another new pi4, the problem is fixed. The hardware causes failure issue. Nevertheless, it is unstable on usb3 in pi 4, sometime still crash.

0 Kudos
Highlighted
Employee
59 Views

Hi User15891561295748856269,

 

Thank you for your reply. Could you please answer the following:

 

  • Have you tried using a powered USB Hub? If not, try it and tell us if it still crashes.
  • Could you please tell us which OpenVINO™ toolkit version are you using?
  • Also, which sample/demo are you using? Is it a Python or a C++ demo?
  • If you can give us more details about how the issue occurs, it would be great. Do you wait a certain time or is it immediately happening after running the demo?

 

Best regards,

David C.

0 Kudos
Highlighted
59 Views

Thanks for your response.

 

This project I tried many times. The outcome is the same.

 

The NCS2 directly plug in USB3 slot. No use any USB Hub aud l_openvino_toolkit_runtime_raspbian_p_2020.1.023.tgz be installed.

 

I run the python code from the below web site and modify the some code by using another inference model, mobilenet v2.

https://www.pyimagesearch.com/2019/04/08/openvino-opencv-and-movidius-ncs-on-the-raspberry-pi/

 

Again, the program be freezed once executed in USB3 slot. I checked it and found out the bug is 'output = net.forward()'. Therefore, I can not use USB3 slot to work while using NCS2.

 

However, once the NCS2 plug in USB2 slot, it is ok and no any error. Totally perfect while the program be executed in USB2 slot on raspberry pi4.

 

0 Kudos
Highlighted
Employee
59 Views

Hi User15891561295748856269,

 

Thanks for the information given. We are currently looking into your issue. Could you please try to run the interactive face detection demo and see if the same error happens? Download the IR files from here. And use this command:

 

./interactive_face_detection_demo -m <PATH_TO_XML>/face-detection-adas-0001.xml -i <INPUT_VIDEO> -d MYRIAD

 

Regards,

David C.

0 Kudos
Highlighted
59 Views

Thanks for your help. The sample can work fine on USB3 slot.

0 Kudos
Highlighted
Beginner
59 Views

I am also looking for suggestions regarding this. Do let me know as well if you find out something useful site.

0 Kudos