Newbie Questions

Do not use MD and SMD. Nobody has found those to work reliably.

Just use IVS set up for human.
 
  • Like
Reactions: Mike A.
@mike32162 In BI5, for the camera ONVIF screen, disable every ONVIF event except IVS like shown.

Also, for IVS, I set memo to a friendly string like "IVS". You can set that to whatever you want, like "person"
cam ONVIF.png
 
Last edited:
  • Like
Reactions: Mike A.
@mike32162 In BI5, for every cam, the default list of ONVIF events should have some IVS rules that should just work for you. Set the memo to a friendly string so you will know which one was triggered when you get alerted.

Also, if you test trigger your camera IVS by walking in front of it, you can then hit the reload button on the ONVIF screen to view the XML for that event. And look at an existing rule to get an idea of how to tweak or create new rule.

Again, stick with IVS only events. And intrusion IVS rules are usually the best (versus tripwire)
 
Last edited:
  • Like
Reactions: Mike A.
@mike32162 In BI5, for every cam, the default list of ONVIF events should have some IVS rules that should just work for you. Set the memo to a friendly string so you will know which one was triggered when you get alerted.

Also, if you test trigger your camera IVS by walking in front of it, you can then hit the reload button on the ONVIF screen to view the XML for that event. And look at an existing rule to get an idea of how to tweak or create new rule.

Again, stick with IVS only events. And intrusion IVS rules are usually the best (versus tripwire)
I'm on the right track. Tripwire seems to be easier (for now). I got intrusion working, but for some reason sometimes the system will not let me complete the box drawing. Have to play with it some more. I also need to go through the calibration procedures as this works better than it did before, but now it's not catching all of the human motion.
 
More questions --
1) I still have some work to do tuning the cameras at night. I've read other posts to figure out where to go from here, but IVS target detection is triggering a cat as a human and setting off alerts two nights in a row. Can I use AI after it passes through IVS for further tuning?

2) Is there a way to disable the inactive profile or force it to behave like others, since recording stops. On the same subject, is there a way to only display defined profiles on the user app? It behaves in sort of a clunky way -- you can restrict profiles to whatever you want, but the app displays all of them. If you click to a profile which is restricted, it changes it to that profile, then will restrict app access, since you've now gone to a profile you're restricted to... I'd rather have the app only display profiles I want users to allow to be changed, and certainly do not want them clicking inactive. The only work around I've been able to figure out is change the hold time to 1 minute so it switches back.
 
Post a screenshot at night with the IVS rules on it for the problematic camera. Did you do the global config for the camera - in many instances that can knock out the false trigger. Or since it is a cat, set a minimum size in the camera IVS that is larger than the cat but smaller than a person.

Yes you can send IVS image thru AI. All Code Project cares about is getting an image to look at - it doesn't care how the image was created.

Have you tried UI3 instead of the app?
 
Post a screenshot at night with the IVS rules on it for the problematic camera. Did you do the global config for the camera - in many instances that can knock out the false trigger. Or since it is a cat, set a minimum size in the camera IVS that is larger than the cat but smaller than a person.

Yes you can send IVS image thru AI. All Code Project cares about is getting an image to look at - it doesn't care how the image was created.

Have you tried UI3 instead of the app?
Ack, forgot to calibrate that one.

For my parents, I'd prefer the simplicity of the app and ease of getting push notifications. Not sure how UI3 would handle it on a phone.
Going to try both calibrating it (I did it for one camera and it was better than not doing it) and possible AI afterward. Now that IVS has gotten rid of the constant requests on a windy day, AI shouldn't be nearly as big of a resource hog.

The other dumb-*** thing I did a few nights ago was left Goggle Chrome open. I woke up to find CPU usage had climbed to nearly 50% and continuing to spike. Disabled everything with BI running and it took an Ah Ha moment to realize what happened. Note to self (and other newbies): If possible run browsers to config your system and search these forums on a different machine.

Other than that, 12 cameras running. 10 are 4MP. I have it set to use 15/15FPS, 8192 CBS on main, D1 1024 on substreams. I noticed in the status tab most of the feeds run near 1,000KBs main and 120kbs sub. Totals are 13,000 kBs main 61 MP/s. CPU is stable below 10%, mostly between 1-2. RAM 1.37 Gig. It's only been up without a reboot for 2 days, so will give this time to settle in.

Edit -- they are all H264, which I presume is still the consensus after reading the forums?

I switched the recording on all the cameras to Sub Continuous + Trigger for purposes of being able to store footage longer.
Next build (likely for myself) will not be Dell Optiplex / SFF. 4 HDD bays + 2 NVME slots & better PSU as a starting point.

If my config would pose an issue down the road, please advise. I wasn't crazy about taking off continuous recording main, but see no choice without more space. It blew through 1TB in a day, now it's averaging ~150GB per day.
 
Yeah, if you want it simple for them and push all in one app, then you need the BI app.

Most of us have gone to the pushover app ($5) that sends us push alerts with images and more ability to easily set things like quiet hours or turn off notifications while out mowing the lawn, etc. and then jump into UI3 if we get an alert image that is something we need to look at live or recorded video for.

H264 is still the most recommended codec.

Yes, Cont+Trigger is more than sufficient, especially at the higher bitrate you are running. That is what most of us use and if Cont+Trigger resulted in poorer performance and ability to know what is going on, we wouldn't use it.
 
  • Like
Reactions: mat200
Did you do the global config for the camera - in many instances that can knock out the false trigger.
Some of these are more difficult to calibrate than others, as I don't have fixed vertical objects to tape off and I'm working alone.
Would it be acceptable to stick a yardstick in a bucket, go to the screen, do the ruler, move it, and repeat the process until I have 3?
Horizontal is easy.
 
Minor question:
I have front door zone set to trigger an alert with a doorbell chime. It works. Wondering if I can do something like have it re-trigger the chime every 15 seconds until the zone is empty.
(Edit -- figured that one out)

Still having issues with the whole thing working at a night. Calibrated zones today, went through the posts on the forum for manually adjusting shutter, gain, etc for day/night, will have to wait until it gets dark to test it out again. I had it working using BI with guidelines you gave (but too many false alarms), but not with IVS at night, yet.

Edit -- & Now I have an issue with a different zone. It triggers once, then no matter how long I wait it won't trigger again unless I restart the camera. It doesn't seem like a BI issue. I walk around with my phone with the camera UI on and can see if/when I trigger the tripwire and it doesn't light up on subsequent attempts.
 
Last edited:
Updates ---
First, much gratitude to @wittaj @looney2ns & everyone else who substantially shortened the learning curve.

I have to wrap this up in a few days & head back home, so I've taken a few shortcuts for now which seem to be stable and working.

I'm having a shit-ton of difficulty dialing in some of the cameras in inside zones at night. The image is there, I can certainly see myself walking around in the dark, but IVS triggering is haphazard at night. I've calibrated it and it works better, especially in daytime, but still have some work to do with a few cams at night. I'm back to the Blue Iris Motion detector in those zones, which seems to work much better. I've installed a Coral AI Edge TPU for AI and it works fairly well (although I have to tweak that). I'm getting detections faster than 60ms, a substantial improvement and it's not bogging down my CPU. Turned off static image analysis and only analysis on motion triggers so it wouldn't constantly ping the server. Here's what I got, in case anyone else is interested:


It's dual edge, M.2 E key. I was not sure if it would work in the wifi slot when I ordered, and since I'm pressed for time, I went with a cheap PCIE adapter card as well.
Almost plug and play. Followed the directions to install the drivers, turned it on, turned off Yolo, and Wala, it works. The single ones I saw on Amazon did not list Windows compatibility, so I did not want to take a chance. This one did.

For the rest of the zones, especially ones in the wind outside, sticking with IVS. Those get enough ambient light at night to not cause the same issues I'm having in a dark garage and other areas.

Next on deck is to lock down the network (which I know how to do) then open it up via VPN before I leave. I tested it with port forwarding and DynDns with wifi off to make sure it works, then locked the port down. Waiting for my router to show up tomorrow so I can finish it.

When I get back home, I'll likely order a few cams and continue this journey (which has now turned into a hobby). I can see how the few Ring cams I have at home are totally over-priced POS compared to this stuff. And... their AI is nothing to speak of. The indoor cam constantly mistakes my 10lb small dog for a human every time it gets up on the couch in front of the camera.

Thanks again for all the help.
 
As an Amazon Associate IPCamTalk earns from qualifying purchases.
  • Like
Reactions: mat200