It looks like you're new here. If you want to get involved, click one of these buttons!
If the Graph runs on the NCS what differences does the CPU you hook up to matter?
Shouldn't a Raspberry Pi perform the same a i7?
Trying to figure out what hardware matters, more sticks or bigger PC
Hi Chicagobob, yes graph runs on NCS. So theoretically any platform supporting Ubuntu 16 should be able to use latest NCS SDK. As far as NCS performance is concerned, it will same whether NCS is connected to a Desktop or a Pi.
My understanding is that though graph runs on NCS, there is other software which communicates with NCS to send/receive data & commands. If underlying paltform can not run that software, you won't be able to use NCS. But I think I ran in to a thread recently, where a particular operation was not supported on Pi.
The PI, although really popular, might not be the best place for this device then. When you think about it the USB data being tied to the network port is not optimal and looking at the one here, it does not have USB3. So the I/O is hampered. I will have to look at ODROID again. I have a few C2's they have the USB on a separate bus although I think its USB 2 as well. Do you know what the advantages of having more than 1 stick really is?
Far as I know each stick runs on a separate graph. So if you have several single image inference to look at you can speed up your process but if your looking at a video stream this isn't helping. Guess I am waiting for the Myriad X stick. Seems like a more logical device. I will work with this one for a while
to get used to the workflow.
Hi, even if you want to draw inferences on a video you are still ok, because you could send each frame from camera/video to a different NCS. So, multiple NCS will speed up the processing. I have not experimented with it, but I think I will plan to do some work around it when I get a bit of free space.
The total time to analyze a frame will not get sped up no matter how many sticks you add.
depending on the graph you will have a built in lag of .5 seconds for multiple object detection.
Everyone wants the MyriadX chip in a stick. Lets hope that comes out at CES 2018;
I hope to bump into the guys from Movidius there.