topic Re: Neural Compute SDK version 2.05 has been released! in Intel® Distribution of OpenVINO™ Toolkit

Neural Compute SDK version 2.05 has been released!

idata — Fri, 15 Jun 2018 23:01:25 GMT

The release notes can be found at https://github.com/movidius/ncsdk/releases/tag/v2.05.00.02

Re: Neural Compute SDK version 2.05 has been released!

idata — Mon, 25 Jun 2018 07:38:35 GMT

Errata: Points 10 and 11 are duplicates in the v2.05.00 release notes.

Re: Neural Compute SDK version 2.05 has been released!

idata — Thu, 28 Jun 2018 08:16:14 GMT

So I took the new API release for a spin today, but when I try to run the hello_ncs_cpp example I get the following errors when opening the device:

_$ make run

making hello_ncs_cpp

g++ cpp/hello_ncs.cpp -o cpp/hello_ncs_cpp -lmvnc

Created cpp/hello_ncs_cpp executable

making run

cd cpp; ./hello_ncs_cpp; cd ..

Can't create semaphore

: Function not implemented

Can't create semaphore

: Function not implemented

E: [ 0] dispatcherAddEvent:533 can't wait semaphore

W: [ 0] dispatcherAddEvent:545 No more semaphores. Increase XLink or OS resources

E: [ 0] dispatcherAddEvent:533 can't wait semaphore

W: [ 0] dispatcherAddEvent:545 No more semaphores. Increase XLink or OS resources

E: [ 0] XLinkOpenStream:909 Max streamId reached deaddead!

W: [ 0] ncDeviceOpen:558 can't open stream

Error- ncDeviceOpen failed

ncStatus value: -2_

What OS resources should I increase and why am I getting : Function not implemented error?

Re: Neural Compute SDK version 2.05 has been released!

idata — Fri, 29 Jun 2018 19:00:06 GMT

Hi,

please update install script to check numpy version : "version need for tensorflow 1.7 numpy >= 1.13.3" > TF not install for this reason. Thanks

…

Building wheels for collected packages: mvnc

Running setup.py bdist_wheel for mvnc: started

Running setup.py bdist_wheel for mvnc: finished with status 'done'

Stored in directory: /tmp/pip-ephem-wheel-cache-1fymcopm/wheels/17/b6/ed/2477b3035457946ce3a1c492dc52a98f6f830d1e33ca7456e1

Successfully built mvnc

tensorflow 1.7.0 has requirement numpy>=1.13.3, but you'll have numpy 1.13.1 which is incompatible.

Installing collected packages: mvnc

Found existing installation: mvnc 2.5.0.2

Uninstalling mvnc-2.5.0.2:

Successfully uninstalled mvnc-2.5.0.2

Successfully installed mvnc-2.5.0.2

…

Installation is complete.

Please provide feedback in our support forum if you encountered difficulties.

The PYTHONPATH environment variable was added to your .bashrc as described in the Caffe documentation.

New bash processes including new terminals will use the updated value of PYTHONPATH.

To use the NCSDK in this shell, set PYTHONPATH now:

export PYTHONPATH="${PYTHONPATH}:/opt/movidius/caffe/python"

USE_VIRTUALENV=yes - used virtualenv in /opt/movidius/virtualenv-python

You must activate this virtualenv to use the NCSDK with python

Activate this virtualenv by entering this command now:

source /opt/movidius/virtualenv-python/bin/activate

pi@raspberrypi:~/workspace/ncsdk $ source /opt/movidius/virtualenv-python/bin/activate

I had to do under virtualenv to let nighly build TF 1.7 install

(virtualenv-python) pi@raspberrypi:~/workspace/ncsdk $ pip3 install --user numpy==1.13.3

Re: Neural Compute SDK version 2.05 has been released!

idata — Wed, 08 Aug 2018 06:12:20 GMT

I upgraded my Virtualbox to the latest available version: 5.2.16 r123759 (including extension pack) and gave API V2 another shot.

After I've re-installed the ncsdk2 (API V2) from scratch, I could finally run the C++ example, but even though it does seem to succeed in communicating with the NCS it still throws up a weird error:

# ~/ncsdk/examples/apps/hello_ncs_cpp/cpp/hello_ncs_cpp
Hello NCS! Device opened normally.
E: [         0] dispatcherEventReceive:200    dispatcherEventReceive() Read failed -1

Goodbye NCS!  Device Closed normally.
NCS device working.

I also ran the SqueezeNet example and it ran fine, however I still got the the error mentioned earlier:

# make run

making prereqs
(cd ../../data/ilsvrc12; make)
make[1]: Entering directory '/root/ncsdk/examples/data/ilsvrc12'
make[1]: Leaving directory '/root/ncsdk/examples/data/ilsvrc12'
LICENSE file already exists

making prototxt
Prototxt file already exists

making caffemodel
caffemodel file already exists

making compile
mvNCCompile -w squeezenet_v1.0.caffemodel -s 12 deploy.prototxt
mvNCCompile v02.00, Copyright @ Intel Corporation 2017

/usr/local/bin/ncsdk/Controllers/FileIO.py:65: UserWarning: You are using a large type. Consider reducing your data sizes for best performance
Blob generated

making run
./run.py

------- predictions --------
prediction 0 (probability 99.12109375%) is n03272010 electric guitar  label index is: 546
prediction 1 (probability 0.335693359375%) is n04517823 vacuum, vacuum cleaner  label index is: 882
prediction 2 (probability 0.151348114014%) is n02676566 acoustic guitar  label index is: 402
prediction 3 (probability 0.0977516174316%) is n03109150 corkscrew, bottle screw  label index is: 512
prediction 4 (probability 0.0661373138428%) is n03532672 hook, claw  label index is: 600

E: [         0] dispatcherEventReceive:200    dispatcherEventReceive() Read failed -1

I had a look at the origins of the alleged Error in the SDK source code and I can pin it to XLinkError_t XLinkInitialize(XLinkGlobalHandler_t* handler) function. What's weird is, the dispatcher seems to be reporting an error communicating with the stick, yet the API calls succeed nevertheless? This is truly kind of puzzling to me.

In search of some answers I tailed the syslog wondering what's going on:

Aug  7 22:18:26 ubuntu-xenial kernel: [  501.346185] usb 2-1: USB disconnect, device number 5
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.758706] usb 3-1: new SuperSpeed USB device number 5 using xhci_hcd
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.778868] usb 3-1: New USB device found, idVendor=03e7, idProduct=f63b
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.778872] usb 3-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.778875] usb 3-1: Product: VSC Loopback Device
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.778877] usb 3-1: Manufacturer: Intel Corporation
Aug  7 22:18:27 ubuntu-xenial kernel: [  502.778879] usb 3-1: SerialNumber: 5CB83DCF74C408
Aug  7 22:18:28 ubuntu-xenial kernel: [  503.973807] usb 3-1: USB disconnect, device number 5
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.530074] usb 2-1: new high-speed USB device number 6 using xhci_hcd
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.659414] usb 2-1: New USB device found, idVendor=03e7, idProduct=2150
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.659424] usb 2-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.659426] usb 2-1: Product: Movidius MA2X5X
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.659429] usb 2-1: Manufacturer: Movidius Ltd.
Aug  7 22:18:30 ubuntu-xenial kernel: [  505.659431] usb 2-1: SerialNumber: 03e72150

I'm not a Linux udevd expert or anything but it seems to me like like the SDK removes the device and re-adds it? Or at least so seems the syslog to suggest.

@Tome_at_Intel you might find some of this helpful to take this further - you mentioned about a month ago you were aware of the issues with API V2, but I haven't seen any bug fixes anywhere so I figured this might be helpful info.

Re: Neural Compute SDK version 2.05 has been released!

idata — Thu, 09 Aug 2018 10:47:02 GMT

@milosgajdos Try switching your Virtualbox USB ports to ohci mode. This change has worked for me.

Re: Neural Compute SDK version 2.05 has been released!

idata — Thu, 09 Aug 2018 20:15:34 GMT

@kixorz I have tried all three options available. See the results below from all but xhci modes fail to detect the USB stick when using API V2:

OHCI:

W: [ 0] ncDeviceOpen:533 failed to find device

Error- ncDeviceOpen failed
    ncStatus value: -2

EHCI:

W: [ 0] ncDeviceOpen:533 failed to find device

Error- ncDeviceOpen failed
    ncStatus value: -2

When XHCI mode is enabled I can at least use the stick with API v2, ignoring the error mentioned in my pervious comment.

Now, I use the latest Mac which only has USB-C ports available so I an adaptor dongle converter from USB -> USB-C. I wonder if that hast to do with the shambles we are seeing all over the place? Maybe something @Tome_at_Intel and the team can look into?

Re: Neural Compute SDK version 2.05 has been released!

idata — Thu, 09 Aug 2018 22:11:09 GMT

this git link replies with error 404

Re: Neural Compute SDK version 2.05 has been released!

idata — Tue, 21 Aug 2018 07:30:11 GMT

@milosgajdos , are you running inside a Virtual Machine hosted on a MAC?

Re: Neural Compute SDK version 2.05 has been released!

idata — Thu, 06 Sep 2018 23:49:21 GMT

Hi,

@AshwinVijayakumar I am issuing the exact same problem on ubuntu (no venv) but I packaged the python code in pyInstaller and am trying to run it.

The code is rnning properly but as soon as it is packaged, this error raises, the device is foundbut can't be open with status = f.ncDeviceOpen(self.handle) == -2

In the best scenario, I'd like to give this code (binary) to a client (this is a PoC) and to be able to run it from a Virtual machine (My client has Windows but no native Ubuntu)

Any suggestion ?

Does any way to use movidius in production mode (i.e. from binary files) is implemented or to be implemented ?

Regards

Magaly

Re: Neural Compute SDK version 2.05 has been released!

idata — Fri, 07 Sep 2018 05:47:13 GMT

The virtual machine is working in the technical part for SDK version where the ubuntu is running error so the device is not working that is going to provide the issue which is netgear tech support will help you to make your work done by providing the process then SDK version 2-05 will be able to release.

Re: Neural Compute SDK version 2.05 has been released!

idata — Tue, 18 Sep 2018 03:57:13 GMT

Work on ubuntu 16.04. (why dont work with 18.04?)

Work on raspberry.

Result process work on float (float16 is finish).

SSD_Object detector run with c++ with coord XY.

Re: Neural Compute SDK version 2.05 has been released!

idata — Tue, 25 Sep 2018 19:05:32 GMT

Hello everyone!

I am currently using sdk v2, and running examples failed. I am using ubuntu 16.04 on Intel Joule.

I was getting the same error earlier, and suddenly it worked once, and started failing again.

I have installed everything based on the scripts given in the sdk. I have also tried installing openCV independently, but the same error persists.

vinay@joule:~/ncsdk-2.05.00.02/examples/apps/hello_ncs_py$ make run

making run
python3 hello_ncs.py;
D: [         0] ncDeviceCreate:307    ncDeviceCreate index 0

D: [         0] ncDeviceCreate:307    ncDeviceCreate index 1

D: [         0] ncDeviceOpen:501    File path /usr/local/lib/mvnc/MvNCAPI-ma2450.mvcmd

I: [         0] ncDeviceOpen:507    ncDeviceOpen() XLinkBootRemote returned success 0

W: [         0] ncDeviceOpen:533    failed to find device

Error - Could not open NCS device.

The device automatically disconnects after the blob gets generated.

Before running the program:

vinay@joule:~$ lsusb
Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
Bus 001 Device 003: ID 8087:0a2b Intel Corp. 
Bus 001 Device 010: ID 413c:2110 Dell Computer Corp. 
Bus 001 Device 009: ID 413c:1010 Dell Computer Corp. 
Bus 001 Device 008: ID 04f2:0939 Chicony Electronics Co., Ltd 
Bus 001 Device 005: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 007: ID 046d:0825 Logitech, Inc. Webcam C270
Bus 001 Device 004: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 002: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

After running the program

Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
Bus 001 Device 003: ID 8087:0a2b Intel Corp. 
Bus 001 Device 010: ID 413c:2110 Dell Computer Corp. 
Bus 001 Device 009: ID 413c:1010 Dell Computer Corp. 
Bus 001 Device 008: ID 04f2:0939 Chicony Electronics Co., Ltd 
Bus 001 Device 019: ID 03e7:2150  
Bus 001 Device 005: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 007: ID 046d:0825 Logitech, Inc. Webcam C270
Bus 001 Device 004: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 002: ID 2109:2812 VIA Labs, Inc. VL812 Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

Any leads would be appreciated.

Thanks!

Re: Neural Compute SDK version 2.05 has been released!

idata — Fri, 28 Sep 2018 03:52:16 GMT

ID 03E7:2150 is your movidius. Usb work.

But Device 019?? it's seems strange.

Your are not device 6 on bus 001.

See on usb driver config. or reinstall os ubuntu 16.04.

Sorry but i work only with c++ compilator.


  #include <stdio.h> 
  #include <stdlib.h> 
  #include <mvnc.h>
 int main(int argc, char** argv)
 {
   struct ncDeviceHandle_t *deviceHandle;
   int loglevel = 2;
   ncStatus_t retCode = ncGlobalSetOption(NC_RW_LOG_LEVEL, &loglevel, sizeof(loglevel));
  // Initialize device handle
     retCode = ncDeviceCreate(0,&deviceHandle);
     if(retCode != NC_OK)
     {
         printf("Error - No NCS devices found.\n");
         printf("    ncStatus value: %d\n", retCode);
         exit(-1);
     }
     // Open device
     retCode = ncDeviceOpen(deviceHandle);
     if(retCode != NC_OK)
     {
         printf("Error- ncDeviceOpen failed\n");
         printf("    ncStatus value: %d\n", retCode);
        exit(-1);
     }
     // deviceHandle is ready to use now.  
     // Pass it to other NC API calls as needed and close it when finished.
     printf("Hello NCS! Device opened normally.\n");
     retCode = ncDeviceClose(deviceHandle);
     deviceHandle = NULL;
     if (retCode != NC_OK)
     {
        printf("Error - Could not close NCS device.\n");
       printf("    ncStatus value: %d\n", retCode);
       exit(-1);
   }
   printf("Goodbye NCS!  Device Closed normally.\n");
   printf("NCS device working.\n");
 }

Compile with :


 g++ run.cpp -o run -lmvnc

(run.cpp is this file, output is run and library is lmvnc)

after compile, ./run

If dont work, reinstall ncsdk -2.05.00.02

Good luck.

Re: Neural Compute SDK version 2.05 has been released!

idata — Wed, 14 Nov 2018 00:19:20 GMT

I get this when running on my Mac. I have run successfully in the past but something has changed that I am not aware of.

I have installed.. ncsdk-2.08.01.02

g++ my_app.cpp -o my_app -lmvnc

$ ./my_app

Can't create semaphore

: Function not implemented

Can't create semaphore

: Function not implemented

E: [ 0] dispatcherAddEvent:570 can't wait semaphore

W: [ 0] dispatcherAddEvent:582 No more semaphores. Increase XLink or OS resources

E: [ 0] dispatcherAddEvent:570 can't wait semaphore

W: [ 0] dispatcherAddEvent:582 No more semaphores. Increase XLink or OS resources

W: [ 0] ncDeviceOpen:527 ncDeviceOpen() XLinkBootRemote returned error 3

Error- ncDeviceOpen failed

ncStatus value: -2

Any suggestions????

Re: Neural Compute SDK version 2.05 has been released!

idata — Sat, 01 Dec 2018 00:50:13 GMT

I tried install ncsdk2 on my mac and my ubuntu notebook.

But when I run hello_ncs_cpp or hello_ncs_py, it failed with error.

How can I resolve this problem?

( neural stick1 has not problem on same machines.)

W: [ 0] ncDeviceOpen:528 ncDeviceOpen() XLinkBootRemote returned error 3

Error- ncDeviceOpen failed

ncStatus value: -2