[Support]: Stationary objects redundant notifications #16942
-
Describe the problem you are havingBefore I start, v.15 has this issue MUCH better and MUCH better debugging tools for it! Thank you. I have garage with glass doors and window. So lighting condition change (when passing car shine light). This triggers redundant notifications as attached. I don't think I can tune motion any better. Car is not moving of course. Below is just a single representative sample. Is there a way to control it? Above is only one representative sample. There are a few more instances like this Version.15 What browser(s) are you using?No response Frigate config fileGarage: # <--- this will be changed to your actual camera later
enabled: true
ffmpeg:
inputs:
- path: rtsp://x:[email protected]/Streaming/Channels/101
input_args: preset-rtsp-restream
roles:
- detect
- path: rtsp://x:[email protected]/Streaming/Channels/101
roles:
- record
motion:
mask: 0.043,0.053,0.367,0.056,0.37,0.093,0.043,0.096
threshold: 40
contour_area: 19
improve_contrast: true
review:
alerts:
required_zones: zoneGarage
detections: {}
zones:
zoneGarage:
coordinates:
0.004,0.138,0.253,0.145,0.297,0.168,0.365,0.138,0.492,0.203,0.561,0.232,0.622,0.228,0.986,0.646,0.991,0.982,0.003,1
inertia: 3
loitering_time: 0 Relevant Frigate log outputna Relevant go2rtc log outputna FFprobe output from your camerana Frigate statsNo response Install methodDocker Compose docker-compose file or Docker CLI commandna Object DetectorOpenVino Network connectionWired Camera make and modelHik Screenshots of the Frigate UI's System metrics pagesNo response Any other information that may be helpfulNo response |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
This is a detection issue, as the model is struggling with this perspective since the default model isn't trained on security cameras. You could try creating an entrance zone and only trigger on those, or consider a better model that may perform better |
Beta Was this translation helpful? Give feedback.
This is a detection issue, as the model is struggling with this perspective since the default model isn't trained on security cameras.
You could try creating an entrance zone and only trigger on those, or consider a better model that may perform better