Dahua IPC-HDW5442T-ZE 4MP Varifocal Turret - Night Perfomance testing

I agree, I could do with the IVS working too. I certainly wouldn't buy any more of these or advise anyone else to buy them while there are such issues outstanding. I'm surprised to see them still for sale.
what is your main problem now, recording on the SD card? What is the firmware build date you are using? The latest new firmware we are testing on it, can be released very soon.Till now no one report on this issue, only on the SMD and smart IR issue, and the new firmware has a big update on these 2 issues.
 
IVs isn't reliable for me especially at night. If I draw triplines in the road (jagged or straight) I might get 20% of the cars that cross the triplines to cause a trigger. IVS boxes don't seem to help either. Perhaps its placement or a lighting issue but it doesn't work at all for me. I ended up turning it off.

The second issue I documented about MD was in this thread: IPC-T5442T-ZE IPC-T5442TM-AS latest new firmware IPC-HX5XXX-Volt_MultiLang_PN_Stream3_V2.800.0000000.20.R.200903 Was this ever addressed??
 
Hey @biggen are you running B&W /IR or color on the ones you’re having trouble with IVS detection?
 
Hey @biggen are you running B&W /IR or color on the ones you’re having trouble with IVS detection?
Color 24/7 for the one in question. Here is a below thumbnail to show the overall lighting of the scene. Its foggy this morning so the image quality is poorer than normal.
 

Attachments

  • Overview 2020-10-28 04.36.49.589 AM.jpg
    Overview 2020-10-28 04.36.49.589 AM.jpg
    556.2 KB · Views: 157
We can ask dahua engineer tomorrow to make a remote checking and support, i send you the latest new firmware to try and the guide how we make remote support.
 
  • Like
Reactions: biggen
On my 5442 bullets, in the examples below the VF model, I'm getting probably 95%-99% accuracy.
FW
VF5442Bullet-FW.jpg

I do think there are some best practices I'm learning with respect to tripwire lines and how they react in the scene.

1) I've found that at the very ends of the tripwire lines, activation can be hit or miss. By extending the lines well beyond the area of interest, there seems to be a better chance of activation.
2) Min/Max object size matters. I've found very small 250x400 or so Min boxes to perform best.
3) There is also a hard to envision (I'm sure an engineer could better explain) "spatial" element to the lines or boxes. I've seen instances where the line was drawn left to right across the street, but at an angle. The subject looks to effectively walk "under" the line without activating the rule. Think in 3D terms not 2D when drawing your lines.
4- Zig Zag lines - A possible benefit to the zig zag line is it provides multiple points of opportunity for IVS to activate.
Picture a normal single tripwire line. IVS has only one opportunity to catch the motion. Now look at my zig zag line. There are 4-5 points of contact on the same “single” line

Here's 3 snaps showing good Human activations (Using AI Human only filter)
human1.jpg human2.jpg human3.jpg

Here are examples where the Tripwire initially missed the subject, but thanks to using an uneven zig zag line, eventually another point on the line activated the rule. The first 2 points of contact on my zig zag dont extend far enough to catch the man in the white shirt, but the 3rd one did.
Again this speaks to A) extend lines well beyond where your eye says they need to be and B) think in 3D terms. The crosswalk was covered, but the human is mostly "Outside" of the crosswalk when viewed in 3D
human4NO.jpg human4YES.jpg
human5NO1.jpg human5NO2.jpg human5YES.jpg

And on a fixed 3.6mm bullet, 2 good IVS events using Tripwire and Intrusion box
human6-fixedlens.jpg human7-intrusion.jpg

And lastly, good activation using AI for vehicles only. Note the tripwire line extends well beyond the street
vehicle1.jpg vehicle2.jpg
 
Last edited:
On my 5442 bullets, in the examples below the VF model, I'm getting probably 95%-99% accuracy.
FW
View attachment 73606

I do think there are some best practices I'm learning with respect to tripwire lines and how they react in the scene.
1) I've found that at the very ends of the tripwire lines, activation can be hit or miss. By extending the lines well beyond the area of interest, there seems to be a better chance of activation.
2) Min/Max object size matters. I've found very small 250x400 or so Min boxes to perform best.
3) There is also a hard to envision (I'm sure an engineer could better explain) "spatial" element to the lines or boxes. I've seen instances where the line was drawn left to right across the street, but at an angle. The subject looks to effectively walk "under" the line without activating the rule. Think in 3D terms not 2D when drawing your lines.
4- Zig Zag lines - A possible benefit to the zig zag line is it provides multiple points of opportunity for IVS to activate.
Picture a normal single tripwire line. IVS has only one opportunity to catch the motion. Now look at my zig zag line. There are 4-5 points of contact on the same “single” line



On my 5442 bullets, in the examples below the VF model, I'm getting probably 95%-99% accuracy.
FW
View attachment 73606

I do think there are some best practices I'm learning with respect to tripwire lines and how they react in the scene.
1) I've found that at the very ends of the tripwire lines, activation can be hit or miss. By extending the lines well beyond the area of interest, there seems to be a better chance of activation.
2) Min/Max object size matters. I've found very small 250x400 or so Min boxes to perform best.
3) There is also a hard to envision (I'm sure an engineer could better explain) "spatial" element to the lines or boxes. I've seen instances where the line was drawn left to right across the street, but at an angle. The subject looks to effectively walk "under" the line without activating the rule. Think in 3D terms not 2D when drawing your lines.
4- Zig Zag lines - A possible benefit to the zig zag line is it provides multiple points of opportunity for IVS to activate.
Picture a normal single tripwire line. IVS has only one opportunity to catch the motion. Now look at my zig zag line. There are 4-5 points of contact on the same “single” line


Excellent! Just changed this and made a huge difference on my tripwire for entering my sidewalk area!
 
  • Like
Reactions: bigredfish
On my 5442 bullets, in the examples below the VF model, I'm getting probably 95%-99% accuracy.
FW
View attachment 73606

I do think there are some best practices I'm learning with respect to tripwire lines and how they react in the scene.
1) I've found that at the very ends of the tripwire lines, activation can be hit or miss. By extending the lines well beyond the area of interest, there seems to be a better chance of activation.
2) Min/Max object size matters. I've found very small 250x400 or so Min boxes to perform best.
3) There is also a hard to envision (I'm sure an engineer could better explain) "spatial" element to the lines or boxes. I've seen instances where the line was drawn left to right across the street, but at an angle. The subject looks to effectively walk "under" the line without activating the rule. Think in 3D terms not 2D when drawing your lines.
4- Zig Zag lines - A possible benefit to the zig zag line is it provides multiple points of opportunity for IVS to activate.
Picture a normal single tripwire line. IVS has only one opportunity to catch the motion. Now look at my zig zag line. There are 4-5 points of contact on the same “single” line

Here's 3 snaps showing good Human activations (Using AI Human only filter)
View attachment 73603 View attachment 73604 View attachment 73605

Now here are examples where the Tripwire initially missed the subject, but thanks to my using an uneven zig zag line, eventually another point on the line activated the rule. The first 2 points of contact on my zig zag dont extend far enough to catch the man in the white shirt, but the 3rd one did.
Again this speaks to A) extend lines well beyond where your eye says they need to be and B) think in 3D terms. The sidewalk was covered, but the human is mostly "Outside" of the sidewalk when viewed in 3D
View attachment 73609 View attachment 73610
View attachment 73612 View attachment 73613 View attachment 73614

And on a fixed 3.6mm bullet, 2 good IVS events using Tripwire and Intrusion box
View attachment 73615 View attachment 73616

And lastly, good activation using AI for vehicles only. Note the tripwire line extends well beyond the street
View attachment 73617 View attachment 73618
I'll give it another go tonight taking your recommendations into account. Andy also send me the newest FW. Hopefully between the two I'll have better luck.

Thanks!
 
  • Like
Reactions: bigredfish
All, apologies for the delay but I wanted to thoroughly test this 10-19 FW. I completed this last night and am posting the results of my findings plus an update video here and in the T5442 FW thread.

SMD & SmartIR Findings:

SMD

  • SMD is rock solid in my testing, best implementation of this yet and looks like Dahua did adjust code according to feedback provided which has led to a huge improvement. As stated in my video (this video focused on SmartIR) I’ve had dogs, cats, raccoons, deers all be ignored and only People & Vehicles captured correctly.
  • As I also mention in my video, if you are still seeing issues with SMD then definitely check out your configuration and also please re-read my statement I shared before about how sensitivity has been programmed by Dahua for SMD. Specifically REMEMBER: SMD at Low Sensitivity actually instead operates as Low Threshold which is very different. Therefore if SMD is not identifying enough targets then choose LOW, if its picking up too many targets (incorrect targets) then MED or HIGH

SmartIR
  • SmartIR still needs work. There are improvements over 9/3 FW but still more needs to be done. Specifically NEAR IR is the issue and needs to be reduced further as FAR IR adjusts perfectly. Take a look as I walk up to camera in the video and you will see that in almost all cases FAR IR is about spot on, NEAR however never adjusts anywhere as close to the same amount especially when the target is center of cam but at an angle. I’ve fed this back to Dahua & Andy, pointed to code and will also link them to the video

Other Notes:
  1. Quicker IR response seen on screen and adjustments in code within this version
  2. When scene has additional light (background for example) the camera does a better job of balancing visible to IR within the scene. Checkout 5m mark and you will see that the floods (2x250w) in the background get kicked down (camera adjustment) as IR attempts to balance foreground subject against huge background brightness and does it well
  3. In IR only mode, fixed exposure, specifically 1/60 has received the most tweaks but only by a small amount over others such as 1/100 and 1/120
  4. When you add a front only light (1x 60w bulb in this case) the fixed exposure of 1/100 or exposure range of 1/60 to 1/100 shows ‘slightly’ better IR response (speed) from previous FW
  5. 1/60 to 1/20 exposure range has also improved when foreground / visible light on target is applied but again just marginally

Summary

The 10-19 FW represents the best release so far in my testing. That is not to say its perfect by any means and the SmartIR issue needs to be fixed in the next release for sure. However, the work that has been completed on the SMD algorithm's here represent a huge difference (positive) over previous releases and the camera also is slightly more snappy, responsive when triggering IVS rules in this version as well.

Lastly, one of the other HUGE benefits that I don't want people to lose sight of here is the addition of manual focus to the 5241-Z12 (also runs VOLT chipset) as part of this FW release that I mentioned HERE (check it out). This is something people have been asking for since the 5231 soft focus issues and really is a game changer. I've tested this and it works really well. Being able to tie to profile is also a bonus.

So while this FW is not perfect yet for the 5442, it has come a long way, brought unexpected benefits (looking at you 5241) and I believe with more work by Dahua on SmartIR specifically, that the next version could be spot on

HTH and as always please let me know with any questions

 
@spammenotinoz Yes thats exactly what its doing (ms) BUT there is an easier way where you don't have to calculate. Little tip in case you hadn't seen it

  • Go into Settings, Conditions Menu
  • Choose Exposure sub menu
  • In the Mode sub menu change to Manual
  • Go down 1 box to the first Shutter selection sub menu and change to Customized Range - You'll now be presented with the 2 entry fields below that, left box represents the lowest shutter/exposure, right box represents highest. These are the shutter value inputs in ms

    To figure out what value to enter (without breaking out the calculator)

  • Click on the drop down arrow in the same Shutter selection menu (where it now says Customized Range) and instead choose (click) your desired shutter speed from the list, lets say 1/2000 then in that same box click back to Customized Range and you will see its pre-populated the number in the right shutter value box below (0.5 in the case of this one). Do the same for the other value you are interested in (it will always display the returned value in the right box, just need to transpose lowest to left, highest to right) and then populate the shutter value fields with those values while leaving Mode to Manual and Shutter to Customized Range
You may have already seen that but if not, maybe that helps, save calculator batteries too ;)

For daytime (again depends on light available in area, time of year etc) but generally having a low of 1/250 (4) and a high of even 100,000 (0.01) for those super bright locations ensures enough speed (even on the low end when it get darker as the day goes on) to capture blur free vehicles / fast moving targets. You can always drop as low as 1/120 (8.33) depending on how long you leave your schedule before you kick into Night config or if its a darker day but don't go lower if vehicles are in your target list otherwise you will be disappointed with a lot of the results.

The other caveat here is of course the purpose of the camera itself. If this is a front door/porch cam that is only interested in capturing people right in front of it then of course you adjust your config accordingly. If on the other hand you are trying to do LPR (with appropriate cam) then you would lock Shutter to around 1/1000 or 1/2000 to ensure crisp plate captures (depending on FOV of camera, angle height distance to / and speed of target etc). So again adjust accordingly but LPR would be an area where you set a specific shutter and then leave it rather than have a value range. The reason for not wanting the value range in LPR is you don't want the camera to 'think' about adjusting exposure/shutter within a limited capture window (FOV) on a fast moving target therefore by locking it down you remove the option for the camera to choose :)

Manual is absolutely the way to go (STAY AWAY FROM AUTO) and I advocate for it wherever I can. I also use this in every system I am involved with as you truly dial in your shutter/exposure for a given situation with a range that can adapt to the location, lighting and target type you are needing to capture.

HTH, let me know with any other questions
Has the left/right exposure boxes changed with newer cameras? I have a 2431 that I'm trying to do this with on a cloudy rainy day, might be the problem maybe. But when I do this transpose trick and switch the numbers for lowest left highest right the image get way too light. However if I keep the boxes like this, 0-0.5, (high left low right) the image shows at a good exposure. But this is a high on the left and low on the right in this example. I'm confused lol. Could this have changed?
 
what is your main problem now, recording on the SD card? What is the firmware build date you are using? The latest new firmware we are testing on it, can be released very soon.Till now no one report on this issue, only on the SMD and smart IR issue, and the new firmware has a big update on these 2 issues.
IVS not working. SD storage wrap-around not working.

We've been talking about new firmware for the last 3 months. At some point (already passed for me) people people start losing faith. It's pitchforks and torches next. :P
 
IVS not working. SD storage wrap-around not working.

We've been talking about new firmware for the last 3 months. At some point (already passed for me) people people start losing faith. It's pitchforks and torches next. :p


IVS (tripwire) definitely working for me. Very well..
 
  • Like
Reactions: looney2ns
latest firmware - 10-19 FW using NVR.

You setup IVS with SMART pSS?
I am still on "V2.800.0000000.16.R, Build Date: 2020-05-29" as I have been waiting to hear the "Good to go" on any other version.

I use the IVS on the camera. I'm not sure if "IVS" per se is available anywhere other than cameras but I assume surveillance software can do the equivalent.
 
I am still on "V2.800.0000000.16.R, Build Date: 2020-05-29" as I have been waiting to hear the "Good to go" on any other version.

I use the IVS on the camera. I'm not sure if "IVS" per se is available anywhere other than cameras but I assume surveillance software can do the equivalent.


try the new firmware and let us know how it goes.
 
I am still on "V2.800.0000000.16.R, Build Date: 2020-05-29" as I have been waiting to hear the "Good to go" on any other version.

I use the IVS on the camera. I'm not sure if "IVS" per se is available anywhere other than cameras but I assume surveillance software can do the equivalent.
Are you recording to a Dahua NVR or to an SD card installed in the camera for the IVS triggers?