coral-pi-rest-server VS double-take

They were completely sold out for 2-3 years (2020 onwards), and Google wiped documentation (https://coral.ai/products/accelerator/ redirects to the main page, which has no reference to the original Coral). I can't tell where there's an official place to buy this. I see some on Amazon, but that might be resold.

OP is probably using an AI accelerator like this: https://coral.ai/products/accelerator which works great on a PI and uses very little power. It will do the Yolo part, but you can't really expect it to do the multimodal LLM part, although you could try to run Florence directly on the PI too.

That said, you can always look at something like https://coral.ai/products/accelerator/ to help with the performance you need.

Should only be ~$60 https://coral.ai/products/accelerator/

The Coral USB accelerator might be of interest if you want to do some light ML with a low power budget.

However, a pi doesn't have the strength to run something like Llama.cpp, of course, so I've been considering using something like the Coral USB Accelerator (https://coral.ai/products/accelerator). As I've been learning more about it, it seems to be very geared towards TensorFlow Lite models. But whisper.cpp and Llama.cpp use ggml models.

# Learn more at https://github.com/jakowenko/double-take/#configuration

I looked at Double Take (UI that lets you do the training of your face recondition easily) and found CompreFace as one of models they support. It looks like what I need but there is a catch... no OpenVino (intel CPUs AI accelerator) support. I really like my low power NVR setup and would like keep it that way. Running AI on CPU without acceleration is both power inefficient and much slower. I have a spare low end GPU but if dump it in the system the current AI acceleration brakes... (I know I can prob fix it but that is a rabbit hole I would prefer to avoid).

CompreFace can be used "in addition" via Double-Take but that is additional detection based on frigate object events.

Not OP, but I _love_ talking about this stuff, so you're stuck with me. :D

I'm a big fan of this project:

https://frigate.video/

It's open source, and you can hook it up to a Coral (or some other things, I think) to get crazy-fast classifications. But CPU is fine for only a few cameras.

Once you get something like that setup, it's just a matter of finding cameras that support RTSP. You get them setup however you like (but preferably wired, with PoE), point Frigate at the RTSP stream, and that's it. Now you've got home security footage that never leaves your house. You can set up a VPN to watch the feeds from elsewhere. Frigate supports MQTT as well, so you can hook it into Home Assistant to get notifications, and even pipe person events into something like Double Take to get face detection:

https://github.com/jakowenko/double-take

have you checked out double take?

I purchased this device off eBay the week of July 4th, and willingly paid scalper prices with the idea that I'd spend much of that week learning Frigate, double-take, and DeepStack; turns out, I had the ML-powered facial recognition, alerts and Home Assistant automations up and running in an hour.