I worked with building and supporting video surveillance servers some years back, and at the time having the system following a person from camera to camera as they walked around was considered advanced. How the times have changed! The servers I worked with could handle up to about a hundred cameras simultaneously, and the largest installation I worked on had over two hundred cameras covering a warehouse, loading bays, a couple of parking lots and so on. That was using three servers and the cameras were all using the then much hyped FullHD (1920 x 1080) resolution. When the installation was done they were amazed that they could actually see who was walking around. With their old system all they got was a blurry "someone" doing something at a frame rate of slideshow... I can't help but wonder what the software is like today. They will have had to move with the times obviously, and for the money the customers were paying for the licenses they better be good.
Good explanation of the fundamentals. I'd add to that the "out of the ordinary" where for example the front of the business faces a street which is busy at some times of the day, and quiet at others or at night, but there is occasionally someone hanging around after calling a taxi or waiting for a friend - no big deal until that person hanging around in front of you business is the same one three or more days (or particularly nights) in a row. Now it is worth logging and saving the video because if you later get a break-in, chances are it was them who was casing the place and waiting for the area to be quiet enough for them to risk forcing entry, likely having first found a way to blind the camera. A good AI system can learn what is normal over time so that it can log and save video of anything which is not normal. That will require it to be set to learn at the start (or it will drive you mad by alerting you for everything), but after a week or two can be set to only add things to "normal" when they have been manually viewed and cleared. It does need more processing power (about a Raspberry Pi's worth per camera), but it will pay off in reduced security staff time once it has built up a good database of the normal activity for the area it monitors. Is it worth mentioning that the storage for such a system should be both in the last place a burglar would look and with a mirror to a secure cloud account in case they do find the onsite copy? It also means that completely normal things can be deleted after a fairly short period, and that saves storage space.
Thanks for your video. I am interested in Synology DVA systems, but the DVA1622 is not worth the money they are asking for, and with only a max of 2 DVA tasks, my interest in purchasing is even less. Then they have the other extreme (DVA3221) for $2,500 with a maximum of 12 DVA tasks. The DVA1622 should be priced at a lower range, and the mid-range version needs to be added with more CPU power than the DVA1622 and more tasks permitted, but at the current DVA1622 price model. I have read a fair amount of reviews from Amazon to online vendors, and most people either regret spending money on the DVA1622, keep it but don't like its performance, or have problems with the unit.
Finding your reviews/tutorials so informative, thanks .. but you never seem to mention the size or make of the discs you are kitting the NAS out with .. can you share please
I worked with building and supporting video surveillance servers some years back, and at the time having the system following a person from camera to camera as they walked around was considered advanced. How the times have changed!
The servers I worked with could handle up to about a hundred cameras simultaneously, and the largest installation I worked on had over two hundred cameras covering a warehouse, loading bays, a couple of parking lots and so on. That was using three servers and the cameras were all using the then much hyped FullHD (1920 x 1080) resolution. When the installation was done they were amazed that they could actually see who was walking around. With their old system all they got was a blurry "someone" doing something at a frame rate of slideshow...
I can't help but wonder what the software is like today. They will have had to move with the times obviously, and for the money the customers were paying for the licenses they better be good.
Good explanation of the fundamentals. I'd add to that the "out of the ordinary" where for example the front of the business faces a street which is busy at some times of the day, and quiet at others or at night, but there is occasionally someone hanging around after calling a taxi or waiting for a friend - no big deal until that person hanging around in front of you business is the same one three or more days (or particularly nights) in a row. Now it is worth logging and saving the video because if you later get a break-in, chances are it was them who was casing the place and waiting for the area to be quiet enough for them to risk forcing entry, likely having first found a way to blind the camera.
A good AI system can learn what is normal over time so that it can log and save video of anything which is not normal. That will require it to be set to learn at the start (or it will drive you mad by alerting you for everything), but after a week or two can be set to only add things to "normal" when they have been manually viewed and cleared. It does need more processing power (about a Raspberry Pi's worth per camera), but it will pay off in reduced security staff time once it has built up a good database of the normal activity for the area it monitors. Is it worth mentioning that the storage for such a system should be both in the last place a burglar would look and with a mirror to a secure cloud account in case they do find the onsite copy? It also means that completely normal things can be deleted after a fairly short period, and that saves storage space.
Great info. I didn't realize you also covered security cam stuff. Now I'm going to have to check out your camera/surveillance playlists!
Thanks for your video. I am interested in Synology DVA systems, but the DVA1622 is not worth the money they are asking for, and with only a max of 2 DVA tasks, my interest in purchasing is even less. Then they have the other extreme (DVA3221) for $2,500 with a maximum of 12 DVA tasks. The DVA1622 should be priced at a lower range, and the mid-range version needs to be added with more CPU power than the DVA1622 and more tasks permitted, but at the current DVA1622 price model. I have read a fair amount of reviews from Amazon to online vendors, and most people either regret spending money on the DVA1622, keep it but don't like its performance, or have problems with the unit.
They want you to buy deep learning cameras such as theirs or Hikvision's. In such case there are more deep analysis that can be concurrently done.
Great explanation, very easy to follow and answered many questions. Thank you!
Great explanation. Thank you
Finding your reviews/tutorials so informative, thanks .. but you never seem to mention the size or make of the discs you are kitting the NAS out with .. can you share please
Thank you !
Does the DVA AI work if the actual camera does NOT HAVE the capability to do AI, as in they are older cameras without the functionality?
I was wondering. You explained.
Thank you 👍
Thanks for the kind words man
Can a cheap NAS run with something like Jenkins from nvidia be a good combination for better AI performance?
Good explanation
So get a DAV and some Reolink cameras?
Ool
What is that stuff to your left?
Your entire video is pointless if you don't explain what that black box is (that you are comparing to) .