I'm using one of these Coral TPUs on one of the micro HP 800 PCs that has two NVME ports and a SATA port. Truthfully, the BI server is running as a Proxmox VM on that machine with the Coral TPU set as a PCI-E passthrough. It runs the same speed as my old dedicated Windows PC i5 setup with an Nvidia GTX960 running the AI. All less than 20ms detection time on 5 cameras. Just be very cautious which Coral TPU you get to match the PCI-E slot for your NVME. I did have to get the Single-TPU model as no Multi-TPU was available for my NVME slot.
Ive been trying to get my tpu to work for about 6 hrs now and no dice. It just keep saying no object detected. And when I change to yolo8 or anything it stops detecting the tpu all together
I tried that on a NAS but was unreliable., maybe processor wasn’t fast enough. I bought a cheap dell mini pc and coral and use that just for surveillance. Works well.
I'm using one of these Coral TPUs on one of the micro HP 800 PCs that has two NVME ports and a SATA port. Truthfully, the BI server is running as a Proxmox VM on that machine with the Coral TPU set as a PCI-E passthrough. It runs the same speed as my old dedicated Windows PC i5 setup with an Nvidia GTX960 running the AI. All less than 20ms detection time on 5 cameras. Just be very cautious which Coral TPU you get to match the PCI-E slot for your NVME. I did have to get the Single-TPU model as no Multi-TPU was available for my NVME slot.
Ive been trying to get my tpu to work for about 6 hrs now and no dice. It just keep saying no object detected. And when I change to yolo8 or anything it stops detecting the tpu all together
How does the ai analysis look overlayed on videos? Are the boxes faster and more accurate?
Why not HomeAssitant eith strong processor and use lets say frigate nvr and llm or chatgpt instead coral?
I tried that on a NAS but was unreliable., maybe processor wasn’t fast enough. I bought a cheap dell mini pc and coral and use that just for surveillance. Works well.
i also have Frigate and the Coral chip on HA for redundancy, works great as well.
Faster than an old GTX 1060? Lower energy consumption?