Came here just for this topic, sort of.
I am using a local CPAI instance using Mike's Yolo11 models as always, but I see
Blue Iris, CPAI and the ALPRYOLO11NET all using my GPU.
Does BI use the GPU for motion detection? It's also using a ton of CPU cycles now as well.
I figured if I told BI to use the local CPAI that only CPAI would be using the GPU, also the ALPR module, but I'm surprised BI is still using both lots of CPU and GPU with CPAI being used for object detection.

Hard to get a snapshot showing them both but they both spike CPU and GPU quite a bit, the CPU is an i9-9900k and the GPU is a 12GB 3080RTX. Back on BI5 with Mike's YOLO11 modules these would always be at 30% or less, the CPU regularly hits 99% now and the GPU is hovering around 60-70% usage.
Is this expected or is the onboard AI not turning off when you tell it to use CPAI?
EDIT: So I did some reading and it turns out use HW acceleration was the problem. I turned it off across the board (since I have substreams enabled) and everything dropped back to normal levels.