Closed cliffjao closed 2 years ago
This most often happens when there is a single frame when the object is missed and something in the zoned area is detected as a car. With all those shadows from the tree and the dynamic contrast for motion detection, it's probably running detection much more often. I would recommend setting some min/max area filters for car on that camera.
Also, note that presence in a zone is determined based on the bottom center of the bounding box, so that zone is considerably larger than it needs to be for the car object. You could also look at adding an object filter mask (again based on location of bottom center of the bounding box) for car to eliminate false positives in areas that should never have a car.
Thanks for the info. Is there any way to save a debug image that shows the moment the object detection box enters the zone? That'd help figure out what exactly is going on here.
The best way to do that is by logging and storing the mqtt messages from the events topic and looking at the message where the zone is listed in the current_zones field.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Describe the problem you are having
Ever since upgrading to 0.10.0, I'm seeing a lot more false positives where an object that never enters the zone shows up as having entered the zone.
Here is an example of a false positive:
That false positive registered for my Driveway zone, which is the bottom-most zone in this screenshot:
Also note that the object passed through the Street zone (one above in the middle) but failed to register for that:
Version
0.10.0-BFECEE9
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
No response
Operating system
Proxmox
Install method
Docker Compose
Coral version
USB
Network connection
Wired
Camera make and model
Lorex E891AB
Any other information that may be helpful
No response