'Run inference on NCS using the API' fails

[idata]

This thread refers to ''Step 4: Run inference on NCS using the API' within the quickstart guide for the NCS.

Up to this point I followed all instructions and got no obvious error messages. I get the following output while executing ''ncs-fullcheck':

~/workspace/mvncsdk/ncapi/c_examples$ ./ncs-fullcheck -l2 -c1 ../networks/AlexNet ../images/cat.jpg

Found stale device, resettingDevice 0 Address: 2 - VID/PID 03e7:2150Starting wait for connect with 2000ms timeoutFound Address: 2 - VID/PID 03e7:2150Found EP 0x81 : max packet size is 512 bytesFound EP 0x01 : max packet size is 512 bytesFound and opened devicePerforming bulk write of 825136 bytes…Successfully sent 825136 bytes of data in 39.059463 ms (20.146488 MB/s)Boot successful, device address 2Found Address: 2 - VID/PID 040e:f63bdoneBooted 2 -> VSCOpenDevice 2 succeeded../networks/AlexNet/graph not foundDevice closed, rc=0

Am I doing something wrong or did I miss anything?

To quote the instructionof step 4: ''[ncs-fullcheck] loads the binary graph of the specified network generated in the previous step […]'.

Does this refer to step 3?

Are the binary graphs created during the execution of the examples or is their something missing?

[idata]

@Fugu_Fish It looks like AlexNet's graph file is missing from the API's network directory. On the Getting Started page, step 4's comment about previously generated graphs is referring back to step 2 which required the user run setup.sh from the API folder (see https://developer.movidius.com/start/software-setup). Running setup.sh (API) will download various files and compile the graph files for each network (including AlexNet) in the networks folder.

In order to create the graph file, you can try manually creating the graph file yourself using mvNCCompile.pyc from the toolkit or use the command inside of setup.sh from the API as a reference.

[idata]

@Tome_at_Intel I've rerun the whole process and stumbled upon this while installing the API (apparently I missed this during the first run):

After downloading the execution of 'MvNCCompile' terminates with ''Caffe Import Error'.

I checked the PYTHONPATH according to section 2.3.3 of the Toolkit Documentation (https://ncs-forum-uploads.s3.amazonaws.com/ncsdk/MvNC_SDK_01_07_07/NCS_API_1.07.07.pdf) and it is set correctly.

Any ideas on what might have gone wrong?

[idata]

I just found a thread (https://ncsforum.movidius.com/discussion/185/setting-up-api-getting-caffe-import-error) with the same error. Executing ./get_models.sh and ./convert_models.sh separately does the trick for me.