Regarding false positives when using CPAI 2.0.8 with Object Detection (YOLOv5 .NET), I have a 85% min confidence filter for "person" and still the bloody spiders are triggering alerts.
is it worth using a custom model to get less false positives? or somehow adjusting the image scale of whats considered a person?
Like when the spiders right on the lens taking up 70% of the screen and its still getting triggered as a person, that an obvious flag to tell it cant be a person.
it works the other way to, bloody birds!!! their far to tiny to be a person, yet still getting triggered.
Any thoughts on what to do here?
See look at this shit! 90% person, this is crazy....