Hello Community! Just today I got my hands on a beautifully blue Movidius Neural Compute Stick.
In the back of it there is a link pointing to developer.movidius.com/start. I typed it into my trusty Firefox and it loaded, showing me simple instructions, which I followed.
1.1 - get Ubuntu 16.04
1.2 - have a NCS
1.3 - internet connection
Okay so far. Then
2.1 - open terminal
mkdir -p ~/workspace
git clone https://github.com/movidius/ncsdk.git
2.3 - have an espresso
2.4 - check "Setup is complete."
Now it gets interesting
3.1 - plug the NCS
make examples<-- errors during this
Just to make sure, I ran _make examples_ a number of times. Caffe/AlexNet would invariably fail (the other ImageNets too).
The attached .txt files are the terminal outputs for each step. install.txt is the installation log from the second installation. make_1.txt, make_2.txt, make_3.txt and make_5.txt are outputs with the NCS plugged in and make_4.txt is without the NCS plugged in (a bit useless but who knows)
I have ROS Lunar Loggerhead installed and sourced and I know it installs its own OpenCV package. AlexNet uses OpenCV and the Movidius installation had another OpenCV installation. Could this be the problem? How do I diagnose this? The logs didn't say a thing about missing libraries or includes not found. The only hits for "warning" were
/usr/local/bin/ncsdk/Controllers/FileIO.py:52: UserWarning: You are using a large type. Consider reducing your data sizes for best performance
What should I do?
My initial suspicions were right. Doing
sudo apt purge ros-* and
sudo apt autoremove were enough to make the SDK examples work. Too bad I don't have ROS anymore. Maybe installing it on a virtual environment solves the OpenCV conflict? I'll try this later.
@jontzbaker Hi, can you try using NCSDK 2 and see if it works for you? NCSDK2 is on the ncsdk2 branch of the ncsdk github repo. You can clone the repo by using the command:
git clone -b ncsdk2 https://github.com/movidius/ncsdk.