How to extract the middle features

[idata]

I'm ported face recognition with NCSDK. One face spend 104 ms with caffe googlenet.

I tryed to remove last two layer "loss3/classifier" and "prob" in deploy.prototxt, But I got some error from mvNCCompile.

I want to extract the middle features and try another algorithm for classifier.

how to extract the middle features?

[idata]

The error message that I get, when I convert a googlenet that remove last two layer "loss3/classifier" and "prob" in deploy.prototxt .

$ mvNCCompile -s 12 googlenet.prototxt -w bvlc_googlenet_iter_3080000.caffemodel -o face_ftmap.graph

….File "/usr/local/bin/ncsdk/Models/NetworkStage.py", line 910, in assign_remnant_buffers t_res = t.assign_remnant_buffers(net) File "/usr/local/bin/ncsdk/Models/NetworkStage.py", line 900, in assign_remnant_buffers (node.outputDimZ,AttributeError: 'int' object has no attribute 'outputDimZ'