Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Intel NCS Troubleshooting Help and Guidelines

Intel NCS Troubleshooting Guidelines

Are you having issues? Try the following:

NCSDK Installation Related Checklist

Platform and OS Check

NCSDK Version Check

Make Examples Error

Sudo/Root Installation Issues

Caffe Import Error

  • On the Raspberry Pi, a user had success by reinstalling the python decorator. This was done by typing the following command: sudo apt-get instal -reinstall python*-decorator. Reference post: (
  • A user reported that installing scikit-image worked for him. This can be done by typing in the following command: sudo pip3 install scikit-image. Reference post: (
  • A user reported that when he/she only sees the Caffe Import Error when he/she uses virtualenvs. When the user doesn't use virtualenvs, he did not run into the problem. Since virtualenvs are not yet supported, please ensure that you are not using virtualenvs while using the NCSDK. Reference post: (
  • Make sure that the account that is using the NCSDK is the same account that was used to install the NCSDK.
  • Check your bashrc file to make sure your Python path is set.
  • Check your Python path by typing in the following command in your Ubuntu terminal to see if your Python path was set by the install script: “echo $PYTHONPATH”
  • If your Python path was not set, you can manually set your Python path by adding the following line to your bashrc file: export PYTHONPATH=/opt/movidius/caffe/python:/opt/movidius/mvnc/python:$PYTHONPATH

Timeout error on Raspberry Pi 3

NCSDK Neural Network Related Checks

Unsupported Networks

SDK Unsupported layers

  • Check the layers (input, hidden, output) used in your network
  • Some layers may not be supported by the current NCSDK. View the supported layers here: If the layers are not on the supported list, please let us know so that we may accommodate your needs as developers.

>1 Image Inference Not Supported

  • Check the batch size of your network’s prototxt file. Currently for the NCS, we only support a batch size of 1.

SDK Layer Errors – Index out of Range

  • Make sure to remove all layers related to training and testing. This includes data layers that rely on labeled input.

Graph File Size is too big

  • Check the size of the compiled graph file. Compatible graph files should be 320 MB or less.

Caffe Input Shape Construct Error

Still Not Working?

If these suggestions don’t solve your problem(s), please create a forum post with the following information:

  1. System specifications (x86? VM? Dedicated machine? Raspberry Pi? Please include OS also)
  2. NCSDK version
  3. Python version
  4. Installation Error Log
  5. Miscellaneous Error Log with error message(s)
  6. Link to Neural Network for debugging purposes (prototxt file, caffe weights, github, pb files)
  7. Any additional relevant information


  • 3 Comments sorted by Votes Date Added
  • I'm running RP3. Loaded latest Stretch and upon "make install" was told "...Tensorflow is not supported on Raspbian Stretch". So...I went back to Jesse, and was told "...which is not supported by NCSDK 1.09. Please upgrade to Raspbiean Stretch and then proceed to install NCSDK".

    Any suggestions for a workaround - any earlier version of NCSDK that will run on Jesse?

  • @jwmac Hi, you can try upgrading to the latest NCSDK (version 1.10) by visiting (referred to above in the SDK version check section) which supports the latest version of Raspian Stretch. There are also added notes about various SDK issues for the Raspberry Pi in the software release notes.

  • @jwmac the issue with TF on RPi is that we don’t (yet) have an elegant why to install TF on RPi. The version that is available now has some incompatibilities with our toolset which we are updating.

    Long story short, our tools should work if you install RPi 1.3 using source install. Also, if you install and compile and create graph files on a compatible machine (say Ubuntu 16.04 on Virtual Box on your PC), you can copy the graph file over (we have examples for inception with a precompiled graph file on our github) and run your application on RPi.

    We are working on making this more homogeneous by working with TF. Thanks for your patience.

This discussion has been closed.