Vehicle / school bus detection
I’m using SecuritySpy in a residential setting. With two days of school in the books, we’ve already had multiple bus schedule problems.
I realize this is a somewhat strange question, but can SecuritySpy’s vehicle detection detect/identify school buses? And could that be broken out into a separate kind of object for alerting purposes, different from cars or delivery trucks or whatever?
My cameras are aimed at our yard/driveway, but get the edge of the road where the school bus travels. I don’t really use any of the current detection features.
Comments
Yes, there are many school busses in our training data for our vehicle detection algorithm, so I would expect these to be accurately detected. However, there is currently no way to differentiate school busses from other vehicles. Note also that vehicle detection will only work reliably if there are normally no stationary vehicles in view of the camera (SecuritySpy can detect vehicles with a very high degree of accuracy, but cannot currently tell if they are moving or not, so any kind of motion around a stationary vehicle will be detected as a vehicle - this is a known limitation of the current implementation of the algorithm).
Not in front of SecuritySpy at the moment, but I’m wondering if a combination of settings like the following would make sense.
If this would even work, a drawback I’ve found is that having extra copies of a camera really hammers the Intel Mac I’m using.
I have 3 physical cameras. If I duplicate them at all for any other purpose, the Mac tends to overheat and crash. This happens even if the schedule and whatnot is only for a very small window of each day.
Another drawback is that other vehicles also travel the street at those times, of course.
The bus does tend to have a distinctive sound, as well, but there’s no AI audio detection in SecuritySpy. I also haven’t checked if my cameras’ audio even works.
Your scheme may work, although it sounds from your description that the bus route may be quite far from the camera, and therefore will occupy only a small part of the video image. This will make for inaccurate detection. It would be much better to add an additional camera specifically for this purpose, with a varifocal lens so that you can zoom into the exact area where the bus is expected (and also to frame the area in question so that there are no parked vehicles in the image). Then you will get accurate vehicle detection for this particular location (although of course you will get detections for all vehicles, not just busses).
Yes, adding additional streams does put more load on the Mac - if you are finding that this is a problem for the Mac, the best way to mitigate this is do set the cameras to a lower frame rate. We recommend 10fps for general-purpose video surveillance.
As far as adding duplicate cameras, I was trying to take advantage of what I saw in another thread (which I think was referenced in 2018?):
"If you add a camera twice to SecuritySpy using exactly the same settings, SecuritySpy will detect this and it will pull in just one stream from the camera, and will not count the second instance towards the camera license limit."
My goal was to a) not exceed my license limit, b) not draw more streams from the cameras, and c) make only certain “cameras” (i.e. special purpose duplicates) active at any given time.
I’ve also thought about trying to do this for certain kinds of person detection, with different zones — maybe front door, mailbox, in front of garage doors, etc. Almost every case I can think of only needs to be active/armed at specific times of day.
Do duplicated cameras continue pull a separate feed from the camera? Do they do so all the time, even when not armed/active?
Do duplicated cameras impact CPU/GPU/etc., even when not armed/active?
If I duplicate the cameras and change settings like lowering the frame rate, will that affect the license usage? It appeared to when I switched a duplicated alternate camera to stream 2 or 3 (lower resolution, lower frame rate). It appeared to soak up another license and adversely affect the performance of the host Mac.
Does the feed or CPU/GPU/etc. situation — whether license or performance — change depending on how the camera is configured? My cameras, to the best of my knowledge, are rebranded RaySharp cameras. I’m using the ONVIF profile with RTSP UDP. I think I have the cameras themselves set to use H.265 rather than their default of H.264 on stream 1.
Thanks!
Hi @jaharmi a duplicate camera will not duplicate the network data stream and will not count towards the license limit, as long as it uses the exact same device settings. This means that the second camera instance can't be set to a different input or lower-resolution stream that the first instance: the device settings must be identical.
Resource usage may be increased. You can mitigate this by enabling the options in the General settings to "suspend motion detection when not needed" and "suspend video decoding when not needed". And, when cameras are not needed, make sure they are unarmed and not being actively shown in live video windows.
In terms of camera encoding settings, H.265 is preferable on newer Macs (but H.264 is strongly preferable on pre-2017 Macs). RTSP TCP is generally the best streaming protocol to use, unless RTSP UDP works better for your specific camera (which is unusual but we have seen this a few times). RTSP TCP is labelled simply "RTSP" in the Format dropdown. Besides that, use VBR encoding with a moderate frame rate (e.g. 10fps).