I have had marginal success working around the Dahua cam IVS limitations on the IP4K-T by using the BI based motion detection, which is more primitive just looking for motion, and using AI to confirm the many alerts it produces, using the "wildlife" label in the To Confirm box under camera AI settings. This works for dog, cat, bird and occasionally a possum or deer. Often the wildlife is misidentified as one of the above, so I am concerned that some are probably getting filtered away. It takes some tuning of the BI motion settings to detect smaller items with too many false triggers. Obviously consumes CPU cycles to have BI process the camera images, so will not scale to a large setup, I currently have 4 cameras set this way using OpenVino yolov11m model with face recognition turned off (because I don't need it). I have also played with the new Continuous AI scan as an option instead of BI motion detection, but that is generally not as good so far and more resource intensive. I am also no expert in any of the above, just hacking my way through it.