Hello,
I’m having some issues with Frigate that make its use unpleasant for me. I’m getting multiple false positives with the “in progress” status, which spam me with messages. How can I avoid these false positives? Additionally, I’d like to add a visible timestamp on the photo when sending. How can I display it? (I’ve already managed to display the boxes).
Thank you.
Firstly put motion masks over those areas.
add
timestamp: true
to your camera conifg.
I already have :
CAM-SALON:
ffmpeg:
inputs:
- path: rtsp://xxxx@xxx:554/ch0_0.h264
roles:
- detect
- rtmp
objects:
filters:
person:
mask:
- 275,923,656,928,770,687,789,653,580,474,229,658
- 1601,425,1822,481,1800,734,1574,605
track:
- person
#- dog
detect:
width: 1920
height: 1080
snapshots:
enabled: true
clean_copy: True
retain:
default: 15
timestamp: true
bounding_box: true
timestamp_style:
position: "tl"
format: "%d/%m/%Y %H:%M:%S"
thickness: 2
mqtt:
enabled: True
timestamp: True
bounding_box: True
motion masks are not meant to be used for this, this is highly discouraged in the docs
The default model is trained on the COCO dataset which does not look like security camera examples. Here are a few examples
Frigate offers many filters for reducing false positives and on this camera unfortunately these are similar size and shape to that of a person. You may find better results increasing the threshold a bit for this camera
How can I use it so ?
I don’t understand the question
How should i use it?
motion masks are meant to be used to block out areas that will have motion not due to detectable objects, like clouds in the sky, timestamps, large trees, etc.