Dahua N3wbie , can’t figure out why IVS events won’t record

[Paiberfunk]

I’m new to Dahua , and I must say that the backwards web interface is maddening in many respects, but I think I finally got a handle on it.

Just when I thought I was getting confident understanding the web UI (5, this is my first EmpireTech-IPC-Color4K-T180 cam) , I discovered that the IVS rules that I set up seem to be taking and storing pictures when the events trigger but NOT recording to the local SD card and it is baffling as to why.

These SMD enabled events (human/vehicle) are meant to take pics and to record 30 seconds of video on trigger.

I’ve tried turning just regular motion detection on too and can’t get that to record either . I’ve checked the motion recording schedule and it is on the default settings with events and alarms on 24 hours and general off for 24 hours. The only way I can get any recordings to show up is by turning the general (green bar) on but then of course it records all the time…

Ive RTFM and I’m lost , so I’m turning to the wise minds here for help , as I’m clearly inexperienced with this platform so far.

 

[bigredfish]

MD must be on to use SMD. It’s a layer on top of regular MD

IVS is completely separate and you shouldn’t try to run both IVS and SMD simultaneously. IVS is generally much better

 

[Paiberfunk]

MD must be on to use SMD. It’s a layer on top of regular MD

IVS is completely separate and you shouldn’t try to run both IVS and SMD simultaneously. IVS is generally much better

I’m confused by this, IVS claims that it uses human and vehicle detection as a target filter.. I assume this means that SMD needs to be on to enable this .

currently I have full frame MD on (default), SMD on, and IVS rules that are filtered by target type of human/vehicle . I can see and test the IVS rule successfully following the guide here :

 

[Mike A.]

I’m confused by this, IVS claims that it uses human and vehicle detection as a target filter.. I assume this means that SMD needs to be on to enable this .

currently I have full frame MD on (default), SMD on, and IVS rules that are filtered by target type of human/vehicle . I can see and test the IVS rule successfully following the guide here :

No, IVS and MD/SMD are separate. Turn off MD/SMD and use IVS for human/vehicle detection.

 

[Paiberfunk]

No, IVS and MD/SMD are separate. Turn off MD/SMD and use IVS for human/vehicle detection.

Are they incompatible? Should I not be getting recording on both if both are on ? My question here is why I can’t get them to record

 

[Paiberfunk]

PS: if I understand what you were saying correctly :: if I turn on MD, then motion detection will happen wherever there is a will happen wherever there is a zone for it.
If I turn on SMD on top of that, then it will only happen in those zones if it matches human or vehicle right?

and then if I turn on IVS , it will additionally trigger on IVS rules .

I guess what I’m looking to do is do a combination SMD in some areas and IVS in different areas for this camera.. is this a bad approach ?

 

[wittaj]

We have found that SMD is very poor and will false trigger way more than IVS. It doesn't work as well as you think it will.

As others have said, IVS is the way to go and MD/SMD is really a holdover from cams that didn't have AI and you really shouldn't use it at all if your camera has AI and it shouldn't even be included on cameras with AI.

So if you want false triggers, continue to use MD/SMD lol

 

[Mike A.]

Not sure why it's not recording but that seems independent of either detection triggering since you are getting images saved and you don't have recording even with MD alone. At least that's how I'm reading your post above.

Generally you should be able to get both. And you seem to be if you're getting images saved from IVS triggers. But in practice you typically don't want to be running both since MD/SMD is going to trigger non-selectively first and independent of your IVS human/vehicle filters. i.e., You'll override your filters and get many falses. That's more what we were commenting on.

 

[Paiberfunk]

Not sure why it's not recording but that seems independent of either detection triggering since you are getting images saved and you don't have recording even with MD alone. At least that's how I'm reading your post above.

Generally you should be able to get both. And you seem to be if you're getting images saved from IVS triggers. But in practice you typically don't want to be running both since MD/SMD is going to trigger non-selectively first and independent of your IVS human/vehicle filters. i.e., You'll override your filters and get many falses. That's more what we were commenting on.

I appreciate that, but if I have an area where I just want any human or vehicle in that area, why should SMD be any different than an IVS set up with lots of crossing tripwires to cover the whole internal area of a perimeter?


At least from following the example in the video I cited above, my impression is that if you want to stake out a certain area doing it this way is equivalent to having SMD on in the area? Is this not best practice ?


Regardless I’m willing to play with it one way or the other but my original concern here is why I can’t get any recordings at all.. is there some sort of video guide that I can follow setting up from scratch default to do this to make sure I’m not missing some weird configuration step somewhere? I have found a lot of settings buried in odd places on this camera and I’m worried that I have missed a key setting

 

[wittaj]

I appreciate that, but if I have an area where I just want any human or vehicle in that area, why should SMD be any different than an IVS set up with lots of crossing tripwires to cover the whole internal area of a perimeter?


At least from following the example in the video I cited above, my impression is that if you want to stake out a certain area doing it this way is equivalent to having SMD on in the area? Is this not best practice ?


Regardless I’m willing to play with it one way or the other but my original concern here is why I can’t get any recordings at all.. is there some sort of video guide that I can follow setting up from scratch default to do this to make sure I’m not missing some weird configuration step somewhere? I have found a lot of settings buried in odd places on this camera and I’m worried that I have missed a key setting

You will find out why most here don't use MD/SMD lol....

 

[Paiberfunk]

You will find out why most here don't use MD/SMD lol....

It's not that I don't believe you.. I do.. i'm just seeking to understand the difference.. what's the difference between "detect any human in this area" and " detect a human crossing this tripwire" ? I assume they're using the same "is this human" detection model (in terms of image analysis) , are they not?

Is it not more laborious to draw both a perimeter and internal criss-crossing tripwires than just to paint a motion area? Or am I misusing IVS if that's what i'm doing ?

 

[wittaj]

They are not using the same algorithm.

It really is a holdover from the old cameras prior to AI and is using that technology (pixel changes) and attempted to add human and vehicle to it with SMD but has way more false triggers than IVS that uses AI learning.

 

[Mike A.]

You'll find the longer you work with cams/surveillance stuff that what should happen isn't always what does happen. lol As @wittaj says, SMD just doesn't work very reliably. You'll get lots of falses vs IVS. Also IVS is a lot more granular in what you want to detect where if you need to do that. SMD uses the entire frame. There might be some cases where you may want to use MD instead of IVS to capture whatever but for Dahua and filtering only humans and/or vehicles, just go with IVS. It works very well.

But now we've got you off on other things vs why it's not recording... Back to that, post captures of those screens and maybe someone will see something in your settings.

 

[Paiberfunk]

You'll find the longer you work with cams/surveillance stuff that what should happen isn't always what does happen. lol As @wittaj says, SMD just doesn't work very reliably. You'll get lots of falses vs IVS. Also IVS is a lot more granular in what you want to detect where if you need to do that. SMD uses the entire frame. There might be some cases where you may want to use MD instead of IVS to capture whatever but for Dahua and filtering only humans and/or vehicles, just go with IVS. It works very well.

But now we've got you off on other things vs why it's not recording... Back to that, post captures of those screens and maybe someone will see something in your settings.

So, mysteriously after a camera reboot.. recording is now happening for everything as it should. This is puzzling indeed... but i'll take it as a sign of "reboot first if things are wonky".

For IVS.. is there a recommended "best way" to use it, or is the perimeter+lots of criss-crossing tripwires the standard way ? It looks like you can make a LOT of tripwires in a space.. and it's not clear how many are needed (nor whether the AI waits for the full human to cross the tripwire or just any part of the human).

 

[bigredfish]

So we've been playing with this stuff for a while now. Trust us.
SMD is an AI overlay on old fashioned MD which sucks. While it can do an adequate job in some scenes, its algorithm as mentioned is still based on pixel activation like old MD.
IVS is superior as to false alerts.

Using SMD and IVS simultaneously puts a major tax on the system resources. This happens on two of my cameras when I tried it.
Its called Tearing and its due to the system not being able to process everything simultaneously very well
View attachment Home_ch7_20230510113111_20230510113133.mp4
















Yes IVS needs to be thought through a bit. I like zig zag tripwires, sometimes encompassed within a bounding Intrusion box. If direction is critical, then zig zag isnt the right way to go, instead use individual tripwire lines, multiples with the same direction filter

There is a 3d effect to IVS lines. Be cautious of drawing lines that targets can simply "walk under".
Example:
Lady being shorter and on the higher side of the Intrusion line has not yet set it off. Taller man, on the shorter segment of the line has tripped the alert

 

[Mike A.]

That's one of those "it depends" questions. Varies depending on the view and what you're trying to do. Generally, an intrusion box + tripwires ensures that you'll get most everything. There's not really any significant cost in terms of resources as long as you keep it reasonable (i.e., a box and a couple of tripwire rules) so doesn't hurt to have more and layer them on. Unless you're trying to distinguish triggers things and then you'd want to be more selective. Or you can just draw a box or tripwires and go with a single rule if it does what you want.

A few things to keep in mind... Leave a little space inside the frame before whatever would hit your rule. Generally, it wants to see the perimeter/line crossing (though there's an appears setting for the intrusion boxes that should work). Also, the boxes/lines work in a 3D space. That is, you need to think in more dimensions than just as the lines are drawn. As an example, if you have a cam facing out across a yard and draw a box on the ground (and entirely on the ground) in front of it, then that may not trip when someone sets foot inside that box. You'd need to draw it so that it extends higher in the view off of the ground in a vertical direction in order to capture someone passing through that area. If that makes any sense...

 

[bigredfish]

Primary role of this camera is to catch crosswalk humans. Or any human that is entering/leaving the neighborhood

 

[bigredfish]

Last example

IVS lines pick up woman and PTZ begins tracking. Due to car it loses her temporarily and goes back to its Home preset.
She actually crosses "underneath" an IVS Zag, but the next Zig catches her and PTZ is again activated

This is why I rarely rely on a single IVS line unless directionality demands it




View attachment HOA Rear VPN_ch1_20230329113511_20230329113611.mp4

 

[Mike A.]

Forgot to mention zig-zag lines. @bigredfish's example reminded me. That often works well vs relying a single line. e.g., In case things don't come from an expected direction or the initial crossing gets missed, that gives you multiple points to trigger.