I have tested MotionEye, KerberOS, and Shinobi. Was also looking at Ivideon, Xeoma - basically anything that I can run in Docker and is local only (no cloud).
I believe this rules out Ivideon as it requires an account on their site.
Also I have avoided Zoneminder based on the reviews I have read about its resource usage and saving method.
KerberOS looked decent, but requires a separate docker for EACH camera - and I am looking for software that can view all cam’s in one instance (current plans are for 4 cameras).
I know Blue Iris is very popular, but want to keep using Docker - and don’t want to deal with Windows.
Shinobi and MotionEye are both looking promising.
Shinobi doesn’t seem to like streaming high res, high bit rate - the whole thing lags badly. Also when I add a second feed it doesn’t want to work (camera keeps dying). Lastly I am using H.265 and this streamed okay - but when viewing recordings there was audio only. If I download the recording it is fine, but in Shinobi, there was no video - not useful.
MotionEye I was trying to figure out how to do passthrough recording - this didn’t seem to be very straightforward though… I need to spend more time on it.
My end goal is to do continuous recording (with pass-through). Then send MQTT messages (or direct integration if available) to HA for automation. I would actually love to use object detection instead of motion detection too due to shadows causing significant false triggers.
Shinobi does have object detection capability, but does not come with it - you have to build it an configure yourself. MotionEye I haven’t seen anything for object detection yet (haven’t spent a lot of time on it) but I doubt it does as this is designed for RPi and is very lightweight.
SO - from the savy to the newb - what are people using?
Anyone have a setup that uses object detection?
Any MotionEye or Shinobi experts willing to lend a hand?
I have enjoyed this learning experience so far (first adventure into Linux world, and home automation too!).
I use ZoneMinder. The resource usage in my opinion is quite adequate for what it does. I’m running it on a NUC though. Don’t know how well it performs on other platforms.
What’s wrong with the storage?
Generally what ZM for me did best is being reliable. With other software there was always some problem with specific camera models. I have some that do RTSP, others that work with snapshots. With no other solution I got everything doing it’s job.
I never had a look at object detection though. I’m just accepting, that ZM is good for zones and just keep it that way.
But: I don’t want to hide the fact, that ZM can be a huge pain too. Especially the recent update (UI overhaul) broke some bits for me, and my Apache configuration seems to cause some minor (but inconvenient) problems.
I really wish there was a better (free) alternative. But so far no other software did the job of only really recording when movement is detected better than ZM.
I had read that ZM saves a massive amount of images instead of video - but that is probably an oversimplification of a specific use (motion detection maybe?). I actually want continuous recording - and just want the detection to create alerts. The camera itself will have a motion triggered recording on itself as a backup - but this gets a lot of false positives. There are some other trigger methods I haven’t tested yet though.
I have thought that if the other trigger on the camera works well enough, I might just use this to send alerts (found a script someone created that can trigger on camera’s internal alert) and just use HA to view the stream/recordings.
Generally that’s true. But it just does that when motion is detected. I guess it boils down to what is required. I like the possibility to go far back in time. I have set it to 2 weeks. And 2 weeks of 24/7 recordings for 5 cameras would require a lot of diskspace. Not taking into account the constant load the storage drive would need to handle. After all I don’t need to be able to see what has not happened all night. Instead it might be useful to have a second look at which stranger rang the doorbell to check if I’m on holiday, a week prior to someone breaking into my house.
Of course this depends on the personal requirements. Just want to give insight why I prefer relevant logging over consistent logging.