First is what makes the new stick 8 x faster? Second question can more sticks give you even better inference/detection performance? If you split each send a video frame to one stick and the next frame to another stick for up to 4 sticks will you get better performance? Can it handle HD frames for inference at 30 fps?15 fps? How does inception V2 or V3 fair on this new stick.
Mouser does not have them, so when is the date of sale in the US expected?
Hi Bob. The below document is the product brief. The statement made is "up to 8x", definitely not 8x on every model.
It is definitely possible to use more sticks to produce better inference performance. The code should be modified to use all the connected sticks. Keep in mind however that there is no automatic load balancing for NCS2.
Handling HD frames for inference at 30 fps or 15 fps is model-dependent. NCS2 doesn't enforce a restriction on this.
We do not publish performance specs unfortunately, but you (customers) are welcome to try experiments and share the results.
Since you answered this question, I am going to ask another, is there a synergy with the Intel Real Sense cameras? Any example code? I have a 435 so I thought I would ask.