"Training" mode

[addeej]

I have been using ZoneMinder for a few weeks now, and I have been really impressed with what it can do - nice work!

One thing that has been a little on the irritating side is getting the motion detection to a stage where the number of false positives as a result of clouds, day/night light levels etc. are at an acceptable level.

Something that would save a huge amount of time spent fiddling blindly with the motion detection settings would be some form of training mode. If you could record for a fixed amount of time (containing a few typical events and non-events), and then replay the same recording as a virtual camera input with different settings. This could allow you to tweak all the various motion detection settings using a consistent video template until all real events are caught and as many false positives are ignored.

I hope this makes some sense to someone, it'd make setting up sooo much easier,

Ed

[Baylink]

I have an idea that the easiest way to do this might be to add another mode to monitors called "Train", where you could point it at a stream/sequence/event as you suggest. I'm not sure whether you'd do this by hacking zmc, or at a higher level, but I agree with you that it'd be spiffy as all hell.

[zoneminder]

If you look at some of the web pages and database tables you might notice that training has been in the system for probably a couple of years now. Unfortunately it's never been finished and is disabled. It's a little difficult trying to figure out exactly how it should work as you need some events to run through and indicate whether to ignore or detect, however that only works for reducing the sensitivity and not increasing it (when there would be no events in the first place).

Until I (or someone else) can come up with a definitive spec for this which works in real world scenarios I doubt it'll be finished in the short term.

Phil