If objection detection is ON, buy a coral or a good gpu.
Doods came out before frigate and it was decent. I moved to doods from motioneye. Supposedly it has improved significantly since. I may take another look.
Frigate came out after doods and had better ui and more features. It was hands down better not to take away from doods. I believe it also added support for coral. Frigate is amazing. i just have concern about long term business model but it is worth paying for.
All this said i would try both. Frigate gained huge momentum after it was first released so everyone stopped looking at anything else.
Thanks for your response;
I just bought a “Google Coral TPU USB Accelerator CPU Module”. I’ll experiment with that. If I needed help getting it on the record, can I come back to this?
I don’t understand anymore. if I install ’ edgetpu_runtime_20221024.zip’ then windows sees my coral and then I can select the coral in VM but when I start HA no USB device is recognized. What am I doing wrong here?? Can someone help me further?
I run HA on virtualbox on windows machine Intel nuc. I have been trying to install Google coral all day but I keep seeing in the log of frigate that no Coral was found! I don’t understand what I am doing wrong.
When I look in windows under device manager, I see coral, as soon as I start HA I don’t see it in device manager anymore! When I then look at Virtualbox, it says unknown device under USB connections. See photo and when I start frigate it goes out after 1 minute, in the log I see the following see fig
HUGE
You just plug Coral into a USB port and tell Frigate where to find it:
detectors:
coral:
type: edgetpu
device: usb
I have six cameras on my Intel NUC i3, and without Coral my whole installation turns into the slowest computer in the house.
From the Frigate docs:
Using multiple Coral USB devices with Frigate can significantly improve performance. By distributing the inference load across multiple devices, you can achieve faster detection speeds and handle more camera streams simultaneously. This setup can reduce the inference time and overall CPU usage, making your system more efficient.
Detectors suck resources, so I use them sparingly. I may look for another Coral. When I bought mine form another forum user a few years ago (during Covid) they were remarkably rare.
Why?
First, I do not understand the love of VMs and Containers for a Home Assistant server installation. They add little, if any, benefits at a great cost of a steep learning curve and more COM port and IP problems than almost anyone needs.
You can keep your Windows license intact by simply buying a new M2 SSD, flash HAOS to it, install the SSD and boot the NUC.
Frigate detect is a bit coarse. I have two detections running: person and car. It’s coarse because of limited computing resources. For example, I wanted to put a camera on my deck to learn how squirrels were getting to the bird feeder. I set Frigate to detect “cat”, and it worked fine for my purpose. (Who knew that grey squirrils could fly)? Frigate can’t detect specific people and as far as I know, there is no way to train it.
Why new ssd, does this not work with the SSD in the nuc itself. And how does that work? I have been working with VM from the beginning because I did not know that there is another possibility. Can you explain this to me a bit?
If you, for example, upgrade to a more powerful mini PC for your Home Assistant server, you just put the original SSD back into the NUC and it’s now a Windows computer.