I've been running both a Dahua WizMind NVR &
Blue Iris for the past 5+ years and I've spent an estimated 200+ hours tinkering with Blue Iris over that time and have upgraded both my cameras and NVR and of course stayed on top of Blue Iris changes.
For me, the #1 thing I wanted when I installed my CCTV system was "accurate" motion alerts when someone was approaching my property. This was a bit harder with my specific setup because my property is surrounded by 50 palm trees and a German Shepherd which made it rather complex to disregard false motion alarms, especially on windy days due to the shadows that would be cast on the ground and sway with the wind. I started my project with just an NVR. The NVR was just picking up too many false positives and it would take 10-30 seconds for a notification to hit my iPhone which sorta defeated the purpose of being alerted as someone is approaching your home. Also when I wanted to check my camera feed I found that opening the app, selecting the desired camera, and loading the full-screen feed often took anywhere from 15-30 seconds each time which was incredibly annoying if I noticed someone approaching the house and wanted to view a live feed to identity the subject. So we are talking anywhere from 30-60 seconds from motion event to viewing live feed which in many cases is too long. I wanted faster motion alerts, and less false positives and I wanted to be able to view a live feed quickly from my iPhone... this is when I found Blue Iris.
Blue Iris itself is easy to install, just like any other Windows app. Loading the cameras was originally very hard for me because I had them directly connected to the Dahua NVR and finding the NVR rtsp link to supply into BlueIris took probably 20+ hours to figure out (I had a newer NVR at the time so information wasn't so readily out there) but eventually I got it working thanks to help from this forum. Before DeepStack AI it was simply the built-in Blue Iris motion filters to tinker with which took me a (blind estimate) of 40+ hours to configure in a way that allowed me to cut out most false alerts but also not miss the real alerts and I had the accuracy to maybe 90%, however on windy days the accuracy would drop to maybe 60% due to the palm trees going wild in the wind. I spend countless hours tweaking the sensitivity & threshold settings trying to get the accuracy up but I either set it too tight and missed real events or too loose and get too many false positives. Then when DeepStack AI and now CodeProject AI came into play the accuracy of motion alerts remained around (estimated) 98% even on windy days and both day and night which gave me the results I was really searching for. For the most part, I can receive an alert to my iPhone within 5-10 seconds of the motion event, and loading the live feed may take another 5-10 seconds. So from motion event, AI review, alert via Pushover to viewing the live feed maybe 10-20 seconds (give or take) compared to 30-60 with the NVR.
However, I still utilize my Dahua NVR and set the most sensitive cameras to record 24/7 and I don't really foresee myself dumping the NVR anytime soon for a few reasons. For one, I like having a backup as both the NVR and Blue Iris PC are in separate locations so god forbid something nasty happens there is an increased chance of recovering evidence (Plus PushOver is in the cloud so alert images could survive a house fire). Also, when I had an actual "incident" that required me to extensively review CCTV footage I found that while Blue Iris successfully captured most "motion" events, it was missing some events & sounds that the NVR successfully recorded. (To be fair, I do not set BlueIris to record 24/7 and instead only record alerts) so in this one case where I really needed evidence, my NVR was there to save me with additional footage & audio. Of course, one can make the argument that if I simply had BI set to record 24/7 in addition to the motion alerts I would have not needed my NVR as backup and I couldn't argue against that other than saying I still do love having my cameras record to two separate locations on separate drives providing me some redundancy.
So my recommendation after 5+ years and 200+ hours of tinkering with Blue Iris.....
- Use a basic NVR with 24/7 recording for all your hardwired camera feeds. Preferably buy 3+ drives and use Raid 5. No "need" for fancy AI features here.
- Plug your cameras into a POE switch and NOT into your NVR.
- Use Blue Iris & CodeProject AI for live feeds & accurate motion alerts
- If you have a NAS, set Blue Iris to auto-export motion alerts older than X days to your NAS for long-term storage (Totally optional)
- Keep the NVR and Blue Iris PC in separate locations if possible.
- Battery backup for your POE switch (and NVR) will ensure your cameras continue to record during a power outage.
Now this one still strikes me as totally odd to this very day.... I have two Ring cameras mounted on a palm tree that I primarily use to send notifications to Alexa that someone is approaching my home and these are INCREDIBLY accurate right out of the box and reasonably fast to alert Alexa "Motion Detected at the Driveway" and I can also set it to automatically bring up a live camera feed on my brand new LG C3 TV. I wanna say that in the past 5+ years of having these Ring Cameras, I can't recall a single false alert. I probably missed some "real" alerts, but I've NEVER received a motion alert for trees in the wind or a dog passing by...etc. Why don't I switch to all Ring Cameras.... well because loading the live feed is slow & the solar panel is ugly (luckily it's in a tree so not as visible) and the camera quality is not nearly as good as the 4K wired cameras... but for motion alerts,
they're incredibly accurate right out of the box.