Community
cancel
Showing results for 
Search instead for 
Did you mean: 
idata
Community Manager
328 Views

Must Caffe be installed for movidius SDK to work?

Hi,

 

I would like to use the Movidius to run some Tensorflow object detection models. I have some issues installing the SDK. I cloned the repo but in the make install step, I run into the following error:

 

/usr/bin/ld: /usr/local/lib/libpython3.6m.a(object.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC

 

/usr/local/lib/libpython3.6m.a: error adding symbols: Bad value

 

collect2: error: ld returned 1 exit status

 

src/caffe/CMakeFiles/caffe.dir/build.make:3057: recipe for target 'lib/libcaffe.so.1.0.0-rc3' failed

 

make[3]: *** [lib/libcaffe.so.1.0.0-rc3] Error 1

 

make[3]: Leaving directory '/opt/movidius/ssd-caffe/build'

 

CMakeFiles/Makefile2:272: recipe for target 'src/caffe/CMakeFiles/caffe.dir/all' failed

 

make[2]: *** [src/caffe/CMakeFiles/caffe.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'

 

Installation failed. Error on line 679

 

Makefile:47: recipe for target 'install' failed

 

make: *** [install] Error 1

 

I see it's an error with the Caffe installation and since I don't plan on using Caffe, I went into ncsdk.conf and set INSTALL_CAFFE=no. This gets rid of the error and SDK installs.

 

However, then I tried to install the examples under the tensorflow folder with make run and run into this error:

 

Traceback (most recent call last):

 

File "/usr/local/bin/mvNCProfile", line 36, in

 

from Controllers.Scheduler import load_myriad_config, load_network

 

File "/usr/local/bin/ncsdk/Controllers/Scheduler.py", line 24, in

 

from Controllers.Optimizer import postParsingOptimizations, selectImplementations, streamEverythingSchedule, fixTensors, eliminateNoOps

 

File "/usr/local/bin/ncsdk/Controllers/Optimizer.py", line 25, in

 

from Controllers.Parsers.Parser.Conversion import Conversion

 

File "/usr/local/bin/ncsdk/Controllers/Parsers/init.py", line 3, in

 

from .Caffe import CaffeParser

 

File "/usr/local/bin/ncsdk/Controllers/Parsers/Caffe.py", line 30, in

 

import caffe

 

ModuleNotFoundError: No module named 'caffe'

 

Makefile:30: recipe for target 'profile' failed

 

make[3]: *** [profile] Error 1

 

make[3]: Leaving directory '/home/biqingqiu/ncsdk/examples/tensorflow/inception_v3'

 

Makefile:12: recipe for target 'inception_v3/.' failed

 

make[2]: *** [inception_v3/.] Error 2

 

make[2]: Leaving directory '/home/biqingqiu/ncsdk/examples/tensorflow'

 

Makefile:12: recipe for target 'tensorflow/.' failed

 

make[1]: *** [tensorflow/.] Error 2

 

make[1]: Leaving directory '/home/biqingqiu/ncsdk/examples'

 

Makefile:57: recipe for target 'examples' failed

 

make: *** [examples] Error 2

 

Note I deleted all other examples except for the tensorflow folder. So it seems that even the tensorflow examples have dependencies on Caffe in order to run. My question is: is this true? Is there a way to use Movidius without installing Caffe? (since I have difficulty installing Caffe too). Why must tensorflow examples also depend on Caffe?

 

Any help is appreciated. Thanks!

Tags (1)
0 Kudos
0 Replies
Reply