[Bug?] mvNCCheck ignores -M and -S switches?

[idata]

Not sure about this, but mvNCCheck seems to ignore the -M and -S switches. If I use "-S 4 -M 8" I get exactly the same as output as when I ommit the switches. Proof:

python3 ./mvNCCheck.pyc network.prototxt -w network.caffemodel -S 4 -M 8

mvNCCheck v02.00, Copyright @ Movidius Ltd 2016

USB: Transferring Data…

USB: Myriad Execution FinishedUSB: Myriad Connection Closing.USB: Myriad Connection Closed.Result: (1, 8)1) 6 1.02) 7 0.03) 5 0.04) 4 0.05) 3 0.0Expected: (8,)1) 6 1.02) 3 2.8312e-053) 7 0.04) 5 0.05) 4 0.0 Obtained values

Obtained Min Pixel Accuracy: 0.0028312206268310547% (max allowed=2%), Pass

Obtained Average Pixel Accuracy: 0.00035390257835388184% (max allowed=1%), Pass Obtained Percentage of wrong values: 0.0% (max allowed=0%), Pass Obtained Pixel-wise L2 error: 0.0010009876521337331% (max allowed=1%), Pass Obtained Global Sum Difference: 2.8312206268310547e-05

python3 ./mvNCCheck.pyc network.prototxt -w network.caffemodel

mvNCCheck v02.00, Copyright @ Movidius Ltd 2016

USB: Transferring Data…

USB: Myriad Execution FinishedUSB: Myriad Connection Closing.USB: Myriad Connection Closed.Result: (1, 8)1) 6 1.02) 7 0.03) 5 0.04) 4 0.05) 3 0.0Expected: (8,)1) 6 1.02) 3 2.8312e-053) 7 0.04) 5 0.05) 4 0.0 Obtained values

Obtained Min Pixel Accuracy: 0.0028312206268310547% (max allowed=2%), Pass

Obtained Average Pixel Accuracy: 0.00035390257835388184% (max allowed=1%), Pass Obtained Percentage of wrong values: 0.0% (max allowed=0%), Pass Obtained Pixel-wise L2 error: 0.0010009876521337331% (max allowed=1%), Pass Obtained Global Sum Difference: 2.8312206268310547e-05

[idata]

@Bug80 For mvNCCheck.py, the options -S and -M only come to play when an image is used as input. When using the mvNCCheck.py program with the -S and -M options, try using the -i option to specify a test image for the network you are checking.

[idata]

Well, that's going to be a problem, because I am not using images as input ;) Is there a way to scale the test inputs without using images?

Of course, in the end a user would like to test the network within the input working range. Outside of that range, the network can give strange results but that does not necessarily mean that the network (or the conversion to a graph file) is not working.

[idata]

_bump_ Is there anyone who knows a workaround for this issue, of if a future release will support test input scaling without using a test image?