Open gulo101 opened 2 years ago
would be great idea
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.
Anything we can do to keep this one alive?
I don't think I will approach it this way, but I can pin this issue so it stays open as a reminder.
What about profiles that can be called up with an MQTT message as well as manually within the dashboard?
EDIT: oops realized I already suggested this a few months ago and Blake said detecting IR would be easier.
I'd like to put in a vote for setting the thresholds via MQTT. Then I could have Node-Red/Home Assistant control the timing.
I would support this as a time or IR based approach. I would not limit this to MQTT based configuration though, as I suspect many users only have an MQTT server to keep Frigate happy, and don't actually use it for anything else. I am one such user.
@toxicantidote frigate doesn't require MQTT anymore
I'd like to mention that while MQTT would be nice, IR detection would definitely be my preferred vote for this. I currently have an issue with Frigate detecting insects at night due to IR light as motion. I could play with settings to adjust sensitivity, but then it would impact daytime recording. It would be very nice to at least be able to configure day/night settings and toggle them with the API easily, or even have Frigate detect IR transition and auto-toggle to night settings if configured.
Setting the thresholds via MQTT would be great! I'm having trouble with object recognition at night, even with IR.
I would like to be able to set min_score and threshold (and maybe based on time of the day. Ideally the effective time range would be dynamically calculated based on location-based sunrise/sunset time +- custom offset (or somehow fed from Home Assistant?), but as easier initial solution it may be hard coded times in the YAML file. This would allow to lower the object detection sensitivity at night to generate alerts despite lower confidence level because lower image quality.
Perhaps add something like this: