Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

Multiple model in NCS

idata
Employee
865 Views

Is it possible to get inference from multiple models on NCS?

 

Either sequentially or parallelly
0 Kudos
3 Replies
idata
Employee
522 Views

@gauravmoon, the following posts might help you:

 

     

  1. https://ncsforum.movidius.com/discussion/112/chaining-movidius-sticks/
  2.  

  3. https://ncsforum.movidius.com/discussion/327/executing-multiple-graphs-on-a-stick#latest
  4.  

 

The crux is that we can't yet run multiple graphs simultaneously on a single stick, but you can run them sequentially. i.e. Use device.AllocateGraph and graph.DeallocateGraph dynamically to switch between different graphs. The time to allocate and de-allocate should be in the order of ms, so shouldn't me much of a problem to do this a couple times in the code. If you do want to run multiple graphs simultaneously in your application, refer the links above.

0 Kudos
idata
Employee
522 Views

@AshwinVijayakumar Is that possible to run multiple graphs sequentially on a single stick? based on the link it seems it only works with multiple sticks? I tried to use device.AllocateGraph and graph.DeallocateGraph alternatively on a single stick, but it doesn't work.

 

basically this is what I did

 

graph1 = device.AllocateGraph(blob1) # use graph1 here graph1.DeallocateGraph() graph2 = device.AllocateGraph(blob2) # use graph2 here graph2.DeallocateGraph()
0 Kudos
idata
Employee
522 Views

@ly_ldq Yes, this way of using the stick is not working for now. Probably this will be fixed in next NCSDK versions but currently workaround for this is usage of multiple sticks.

0 Kudos
Reply