Watches an RTSP camera stream for pose-based gestures and fires HA events you can use in any automation. Runs as a Home Assistant Add-on (HA OS / Supervised) or as a standalone Docker container.
1 Like
I installed this and as soon as I start it the CPU usage goes to 100% and stays there. I tried adding yolo_pose_model: yolo11n-pose.onnx to the config but it removes itself as soon as I save the config file.
I run it on a RP5 with only moderate effect on the CPU. I am using the Supervisor OS version of HAAS. I am not sure about the yolo11n attempt - will try to duplicate on my system.
I run HA on a dedicated computer running a Core i7-11700K with 32GB of memory. That is all the computer does
This is my normal load
This is what happens when I start the add-on let it run for a little then stop it.
This is the config I am using
Any updates on this? I have installed all the new releases that you have put out and can set the options now, yet the CPU still goes straight to 100% and stays there no matter what I choose.
The CPU will tend to be very high - it will use as much as it can to get as high a frame rate as possible. That said, on my RP, it has not noticeably affected load average, just CPU (I have four processors on my RP5, and my CPU is at ~300% or so. The rest of HAAS does not seem to register any slowdowns.
I added a 480 model to go with the 320 and 640. It can be selected in the App Configuration page as yolo_imgsz. With 480, my CPU dropped to ~300%. With 320 it will drop to ~200%.
I will test it more. It does seem though that yes even with it reporting 100% CPU it does not seem to effect the speed of anything else.