Create from 0 AI local resource

Hi there ​@bpaolo, welcome to Axelera! 😃 

Hmm, so, looking at where Axelera might fit into your architecture, your "Local AI server" component could use either the Metis PCIe card or Metis M.2 card. Either of these are great for accelerating AI inference locally - particularly computer vision tasks like monitoring lights/sensors/cycles through camera feeds.

We've got experimental LLM support too, but that's early days and only works on the PCIe cards due to the amount of available memory.

In terms of integration with your stack:

• Voyager SDK runs on Ubuntu/Debian, either natively for Ubuntu 22.04 (although native 24.04 is imminent!) or it’s deployable via Docker
• You'd handle MQTT/PostgreSQL/dashboard integration wouldn’t be something Axelera’s gear would contribute to directly - Metis just provides fast AI inference results that your code/stack can use however you need it to

But certainly the edge AI side of things is where Axelera could help! That’s what we’re all about 👍

First of all, thank you for your prompt reply. At this point, I would like to ask you to respond to the issue of the system's poor documentation. Could you please provide me with some references and examples of image processing applications? What I need is to interpret the state of plant leaves. Thank you in advance and I look forward to hearing from you.

Hi Bpaolo,

Which part of our documentation do you find lacking? We’re always happy to hear feedback on how to improve our usability!

I recommend looking at our quick start guide to get you started using our products. Afterwards, since you mentioned using C++, our axinferencenet guide could help you get started setting up your pipeline using C++ APIs.

As for setting up your own pipeline to interpret the state of plant leaves, do you already have a target network in mind? Here is how you deploy your custom model, or if you simply have custom weights for a model from our zoo, try our custom weights tutorial.