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

Inconsistent Output from CPU vs NCS

Hi there,

 

I'm having difficulty getting equivalent results from the ncs and cpu, and was wondering if anyone else has encountered similar issues?

 

Steps to reproduce:

 

     

  1. Take the AlexNet example from the ncappzoo: https://github.com/movidius/ncappzoo/tree/ncsdk2/caffe/AlexNet
  2.  

  3. Run make all, a deploy.prototxt and bvlc_alexnet.caffemodel weights file will appear
  4.  

  5. Take deploy.prototxt and remove everything after the conv1 layer - expected output size is now 96 x 55 x 55
  6.  

  7. Generate a new graph using mvNCCompile
  8.  

  9. Run the graph on the stick, and store the output of 290,440 floats ( use script similar to this: https://1drv.ms/u/s!AosBHN3IHTLWhoJ7Hk3DnpRZMOo31w )
  10.  

  11. Run the model on the cpu, and store the output of 290.440 floats ( use script similar to this: https://1drv.ms/u/s!AosBHN3IHTLWhoJ840l2yBCQlVo-aA )
  12.  

  13. Compare results in something like Matlab, Octave, or Python
  14.  

  15. Output looks like this: https://1drv.ms/u/s!AosBHN3IHTLWhoJ9GmzKhS7sIxurqw
  16.  

 

Environment:

 

NCSDK v2.05.00.02

 

Caffe v1.0.0

 

Ubuntu v18.04 in VirtualBox 5.2.16

 

Many thanks..

0 Kudos
3 Replies
idata
Community Manager
80 Views

@Jay Have you tried compaing the results from the ncappzoo Alexnet example (https://github.com/movidius/ncappzoo/blob/ncsdk2/caffe/AlexNet/run.py) instead of the example at https://1drv.ms/u/s!AosBHN3IHTLWhoJ7Hk3DnpRZMOo31w? I'm curious to know what the results would be. I wonder if it could be a preprocessing issue since none of the appzoo examples use the transpose preprocessing step seen in the python scripts above img = img.transpose((2,0,1))

idata
Community Manager
80 Views

Hi @Tome_at_Intel,

 

Thanks for that - it seems to have fixed things! The new histograms are here:

 

NCS Version:

 

https://1drv.ms/u/s!AosBHN3IHTLWhoMCk3rqG5v6z2-Crw

 

PC Version:

 

https://1drv.ms/u/s!AosBHN3IHTLWhoMBzM8rCJ1h_NztIQ

 

All the best!

idata
Community Manager
80 Views

:)

Reply