I have a tensorflow PB file and mvNCCompile fails creating a graph

[idata]

Now what? -- Background I created a graph using retrain.py from tensorflow for poets

the pb file looked like it build correctly. I also have a label file. So I thought I would try and build a graph. Ran this. sudo mvNCCompile deeplearning_graph.pb -s 12 -o deeplearn.graphIt failed miserably. I no linux or python expert but it looks like I am missingfiles. Anyone have a how to compile this graph?

Traceback (most recent call last):

File "/usr/local/bin/mvNCCompile", line 35, in from Controllers.Scheduler import load_myriad_config, load_network File "/usr/local/bin/ncsdk/Controllers/Scheduler.py", line 25, in from Controllers.Optimizer import postParsingOptimizations, selectImplementations, streamEverythingSchedule, fixTensors File "/usr/local/bin/ncsdk/Controllers/Optimizer.py", line 25, in from Controllers.Parsers.Parser.Convert import Convert File "/usr/local/bin/ncsdk/Controllers/Parsers/init.py", line 3, in from .Caffe import loadNetworkObjects, parse File "/usr/local/bin/ncsdk/Controllers/Parsers/Caffe.py", line 30, in import caffeImportError: No module named 'caffe'

Sorry thought it would just work.

[idata]

Did a quick search and my

echo %PYTHONPATH comes back as:/opt/movidius/caffe/pythonRead as many examples and trouble shooting guides and have no idea why this has an error. From the page https://movidius.github.io/ncsdk/tools/compile.html Using a PB file should be OK.TensorFlow:network.metanetwork.pbAs for the rest I do not even think I got that far. I re-ran the make for example in NCSAPPZOO so now I am at a loss. Are there any more reasons this would not even try to compile

[idata]

@chicagobob123 Strange that you are using a TensorFlow model but you're having Caffe import errors. Anyway, if you are using the Tensorflow for poets from Pete Warden's blog, he uses some TensorFlow ops that we don't support, so if you're using his version, it won't work.

[idata]

Thanks, well now I have several questions.

Since I can't retrain an inceptionV3 graph using retrain.py and use the pb file. Heck I can not even get mvNCCompile to even try. What do I do now? What is the Movidius way to retrain a graph and get good recognition?