delirii Posted January 24 Share Posted January 24 I'd like to use Coral Edge TPU with my Gen8 (https://coral.ai/products/m2-accelerator-dual-edgetpu/). The purpose is to greatly boost objects detection from multiple HD camera streams (running at home). I've started using Frigate (https://github.com/blakeblackshear/frigate), which is really promising compared to many other solutions I've tried before (false positive nightmare). I think the best option is to use Gen8 PCIe slot, using such an adapter (when back in stock 😞 ): https://mbs-shop.online/tproduct/365624859-195484769511-dual-edge-tpu-adapter-pcie-x1-low-profil  Did someone test such a card (single edge TPU, or even dual edge)? Any clue about how to use such cards with a MS Gen8? Link to comment Share on other sites More sharing options...
schoondoggy Posted January 27 Share Posted January 27 As the Coral device is a M.2 E-key device you could mount it on any compatible M.2 PCIe card. The question is can you get the Coral device? They appear to be sold out everywhere. I work for a system integrator, one of our customers designed a vision system for manufacturing, camera inputs to a small formfactor PC. They were planning to use Coral device using M.2 slots on the PC, but due to no availability, they redesign their software and use a NVIDIA Quadro GPU for the AI for their cameras. It is a much more expensive solution, but they could not wait for Coral/Google. I think the low cost aspect of the Coral devices will be very compelling once they work out supply chain. Link to comment Share on other sites More sharing options...
delirii Posted January 28 Author Share Posted January 28 @schoondoggy You're right, it's impossible to buy such a card right now. But as I can wait, I'm just planning this. And even if it takes 6 months, it's OK. I can run the software using the CPU (I've just bought a Xeon 1265). Not ideal, but it works Once Coral cards will become available again, I'll buy one of them.  Beside this, Frigate documentation mentions this: "There is no support for Nvidia GPUs to perform object detection with tensorflow". Nvidia is a big no-go: too expensive, too power hungry, too big to fit in the case I guess, not supported by Frigate... This choice can't be mine definitely, at least not for my current use case.  I was just wondering if anyone here had tested them. Just to know how well they can work in our beloved MS Gen8 🙂 Link to comment Share on other sites More sharing options...
Qba Posted February 22 Share Posted February 22 USB Coral TPU works fine Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now