3 ms·

I have something similar. It's not tracking though. Drawing around 10W on a pi, around 7W on a Jetson.

• 4ggr0 2y ago
  not sure if i'm misunderstanding - you've got a similar GPU to a 1060 hooked up to a pi?
  • lelag 2y ago
    OP is probably using an AI accelerator like this: https://coral.ai/products/accelerator 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.
    • mobilemidget 2y ago
      This works better in my experience, https://hailo.ai https://hailo.ai https://www.raspberrypi.com/news/raspberry-pi-ai-kit-available-now-at-70/#comment-1597156 https://www.raspberrypi.com/news/raspberry-pi-ai-kit-availab...
      • synergy20 2y ago
        coral has pcie module which is 1/4 to 1/3 of the price
  • hcfman 2y ago
    Not a pi. A Jetson. Still an arm SBC though.