Making MobileNet SSD fails on pi

[idata]

Following https://github.com/movidius/ncappzoo/tree/master/apps/security-cam I run into the error:

pi@raspberrypi:~/workspace/ncappzoo/apps/security-cam $ make run

Making MobileNet SSD…

make[1]: Entering directory '/home/pi/workspace/ncappzoo/caffe/SSD_MobileNet'

making caffemodel

caffemodel already exists, skipping download.

making prototxt

Prototxt file already exists, skipping download.

making compile

/bin/sh: 6: mvNCCompile: not foundMakefile:87: recipe for target 'compile' failedmake[1]: *** [compile] Error 127make[1]: Leaving directory '/home/pi/workspace/ncappzoo/caffe/SSD_MobileNet'Makefile:11: recipe for target 'mobilenet-ssd' failedmake: *** [mobilenet-ssd] Error 2

[idata]

I flashed the latest version of rasbian, followed https://www.pyimagesearch.com/2018/02/12/getting-started-with-the-intel-movidius-neural-compute-stick/ to the letter, and still get this error

[idata]

@robmarkcole In the link you provided the user installs in api only mode. If you install in api only mode, you cannot use the toolkit to compile graph files on the RPi. An option you can choose is to install the ncsdk on your laptop, compile the graph files on your laptop or pc and transfer the files over to the Pi. Make sure and put them in the corresponding network directory (i.e. if you compiled mobilenet-ssd’s Graph file, put it in the ncsdk/caffe/mobilenet-ssd directory). You can use this method to generate graph files if you want to install in api only mode.

I recommend using the command “make install” to install the entire sdk to be able to compile graph files on your Pi. If you install the sdk using “make install”, you can compile the graph files directly on the pi.

Which ever method you choose you can use the command “make run” to run the app. If you are using a pi cam, use “make mobilenet-ssd” And then “python3 security-picam” to run the picam version of the app.

[idata]

@robmarkcole If you have issues installing the entire ncsdk in your Pi, you can increase the swap space on the Pi and try again. Thanks!

[idata]

Hi @Tome_at_Intel I ran pi@raspberrypi:~/workspace/ncsdk $ sudo make install and after a long while got the following error:

[100%] Building CXX object python/CMakeFiles/pycaffe.dir/caffe/_caffe.cpp.o c++: internal compiler error: Killed (program cc1plus) Please submit a full bug report, with preprocessed source if appropriate. See <file:///usr/share/doc/gcc-6/README.Bugs> for instructions. python/CMakeFiles/pycaffe.dir/build.make:62: recipe for target 'python/CMakeFiles/pycaffe.dir/caffe/_caffe.cpp.o' failed make[3]: *** [python/CMakeFiles/pycaffe.dir/caffe/_caffe.cpp.o] Error 4 make[3]: Leaving directory '/opt/movidius/ssd-caffe/build' CMakeFiles/Makefile2:1245: recipe for target 'python/CMakeFiles/pycaffe.dir/all' failed make[2]: *** [python/CMakeFiles/pycaffe.dir/all] Error 2 make[2]: Leaving directory '/opt/movidius/ssd-caffe/build' Makefile:127: recipe for target 'all' failed make[1]: *** [all] Error 2 make[1]: Leaving directory '/opt/movidius/ssd-caffe/build'

I put the full inputs and outputs of my session on https://hastebin.com/relinuyowe.java

Please advise.Thanks

[idata]

@robmarkcole I would avoid using “sudo make install”. Try installing the ncsdk again using just “make install”.

[idata]

@Tome_at_Intel I ran make install and after apparently almost completing the make my pi desktop has been frozen for the best part of a whole day at [100%] BUILDING CXX ..., is this normal/recoverable?

[idata]

@robmarkcole This is definitely not normal, but I have experienced a lockup one time during an sdk install. Not sure what causes it (heat, installation crash, or otherwise) since the install was left to run overnight, but the OS was not responsive and I had to unplug my Pi. I plugged the Pi back up and continued running make install and it seemed to continue successfully.

[idata]

@robmarkcole Note: please allow sufficient time for the ncsdk to complete its installation process. The ncs installation should take well over an hour. The longest part is usually building Caffe. Building opencv (which is done manually afterwards by running the install-opencv script) can take over four hours. Thanks!