So the NVR unit itself is a PC, based on their website. Perhaps silly questions, but just to be clear, you have to provide the camera(s), and for each camera you pay up to $800 for "live analytics" and up to another $400 for "video search"? Are there options to record based on specific events only? The analytics work during later playback, yes? So "non live"?
By "video search" they mean their object search module (detection of 50+ objects). The video does not get recorded directly to the Ironyun device, it pulls video from an NVR or VMS (with a specific driver from Ironyun) so if there is a specific event it will grab the clip from the NVR/VMS at the time of the event. The analytics are analyzing the live stream of the camera, not playback.
To clarify, Ironyun calls their device "AI NVR" but it does not function like an NVR, it only saves the snapshots from an event and does not record video to the device. To actually view video from an event it needs to be paired with an NVR or VMS (with a specific driver from Ironyun)
A useful an in depth article. One question, I am struggling to find info about the hardware. How many channels does their AI NVR support or what are the hardware requirements to run it on your own system?
We’re a security integrator and have installed multiple IronYun AI servers supporting several featured analytics in a high camera count enterprise environment. First, we did initially encounter false alarm issues on rain drops. This issue was easily and permanently corrected and has been completely eliminated. We’ve seen consistent results in all weather situations in the high 90 percentile. Second, IronYun does provide an end user interface to many popular VMS platforms. After testing many popular AI solutions we say with confidence IronYun proved to be best of class. We give them an A+.
Dave, congrats on your first comment. What changes did you make to completely eliminate the false alerts on rain? We found that even increasing the confidence needed to alarm on a person wouldn't get rid of the false alarms.
This is Dave from IronYun. We changed the confidence threshold on Person objects to 0.85 and the raindrop problem disappeared. I think IPVM initially used 0.50 as the Person confidence threshold and that may be the cause of the raindrop false alarms. We have seen rain, snow, and ice on many camera lenses without any false alarms for over a year.