Blue Iris

I've got an i7-6700T with a 4gb GTX 960, 16gb ram, and 1tb m.2 storage running blue iris with codeproject. The GPU is handling the object detection but it's surprisingly slow (~300ms +/-) I'm feeling pretty certain on getting the Coral USB accelerator but i saw the m2 accelerator with dual edge TPU (https://coral.ai/products/m2-accelerator-dual-edgetpu).

First off, I don't fully understand the A, E, or M connections so I still need to look into that. But, here's my thought:

If I pull out the M2 1TB and replace with the m2 accelerator I'd have 2x the power of the USB accelerator. I'd use a 2tb SATA SSD for storage and (I think) the R/W speed should still be fine.

Ultimately I'd like to dial in the modelling to allow for accurate facial recognition. I also like the idea of trying to snag license plates and other more accurate vehicle identifiers (FedEX logo for example).

Your feedback is appreciated.

Actually, it looks like the alienware alpha r2 only comes with M key which is incompatible with the E key of the m2 accelerator. bummer.

there are several different form factors for the coral TPU, you can also buy adapters to get the coral pci-e TPU on a regular PCI-E card using your regular motherboard slot

you can check these here: https://github.com/google-coral/edgetpu/issues/256

https://www.makerfabs.com/dual-edge-tpu-adapter.html

The dual TPU pci-e card required pci-e bifurcation, if you do not have that, it will only show one of the TPUs.

Also you can use this one https://www.makerfabs.com/dual-edge-tpu ... m-key.html