IPC-T5442 (4MP) vs. new IPC-Color4K-T?

I would try to use a constant shutter speed instead of a range to see what happens. Like 1/120, and mess around with the speed, then see where you end up with your other settings.
I will give that a try. I am sure the rain and wet ground was not helping either.

Thanks a lot for posting these - nice to see some real world comparisons and I'm sure others will appreciate these. The 4K-X really looks good until you see some motion - I was surprised by the amount of blur. It's almost like everyone needs one 5442, and one 4K color. Do you happen to have any captures from the 5442 with the IR off? The 5442 has a 1/1.8 sensor but half the pixels to spread the light across so I'm curious how usable it is.

I was taking another look at the pics you posted earlier and also really noticed the huge difference in detail that the 5442 gives on the neighbors support post for the front porch. Everything just looks so much better. I wonder what that is? Have you found any tweaks?
Here are two videos from Friday night when I was messing around with the 5442 and keeping it in color. I don't remember the settings I ended up with, but I was using @MikeLud1 's settings guide as a base. I feel this is the best I was going to be able to get and it look presentable and not washed out from too much post adjustment. I am not sure on the neighbor's post. It could be the difference in settings between the two cameras.






So you are starting to see and recognize what we say often:

  • The bright static image at night usually results in motion blur. I can make a cheap $40 camera look like noon out when it is midnight, but the motion is ghost/blur city.
  • Once you dial it in, the image gets darker
  • A great camera in a bad location results in poor performance
  • A fixed-lens camera on a 2nd story will not provide IDENTIFY images no matter how great the camera is. Even if you had ZERO motion blur, you still won't IDENTIFY. It makes a good OBSERVE location though.

So a couple of suggestions.

  1. You need to decide the purpose of this camera. As you are seeing, even with a super fast shutter, you still won't IDENTIFY, so make it an OBSERVE overview location. So slow the shutter down to the point that you can start to make out color of cars and whether it is a two or four door, along with ability to ID clothing color and skin of a person.
  2. The roof is preventing a lot of light from getting to this camera and is inevitably ruining the exposure. The roof is essentially acting like a sun visor on a car, well in this case a light visor from the bottom. Either accept that and run the camera as more OBSERVE and RECOGNIZE or move it down.
  3. Consider a varifocal here like the 5442-Z4E or the 5241-Z12E so that you can OPTICALLY zoom in to a pinch point in order to get the IDENTIFY shot if that is your goal.
  4. Try bumping the gain to 50 or 55 and drop the NR to 40.

The stutter is a limitation in your system - either the cameras are going thru the router and it can't keep up or a peripheral along the way is bottle necked, or the computer cannot keep up.
I agree with everything you said. I had expected the 4K-X to be much better than the 5442 in color low-light (which it is), and had hoped it would do better, but I had not really expected it to do better than Observe. This position's main purpose/intent has always been a general overview angle. I use it to just see what is going on outside, to track where someone came from, and now with the LPRs installed I would like for it to serve as an addition to better identify Make, Model, Color, etc of the car whose plate was read. I have other 5442s closer to the ground for the other uses and am not concerned with Recognize or Identify.

I will make some more changes this week before I decide the fate of the camera and whether I put the 5442 back up. I would rather not move it down to the lower soffit as I already have a lot going on down there and don't want my neighbors to think I am any weirder than they already do, but there may be one spot I could move it to that wouldn't be as much of an overview but I would be able to point it at the street and get details on cars. If I did this I would keep the 5442 on the upper soffit.
 
Yeah, in that case then, I would up the gain of the 4K just to the point that any more would get ghosting and look at slowing down the shutter just a bit. I think that would give you better color than the 5442 trying to do the same thing as an OBSERVE camera.
 
Yeah, in that case then, I would up the gain of the 4K just to the point that any more would get ghosting and look at slowing down the shutter just a bit. I think that would give you better color than the 5442 trying to do the same thing as an OBSERVE camera.
When cars drive by their headlights obviously add a lot more light, which helps a ton.

Also, any idea what would be causing my 1:40 video to get slowed down and end up being 1:55? What's the "proper" way to export clips from BI? I drag the crop pointers to where I want, copy to clipboard, and then export as MP4.
 
Yeah it is amazing how well the camera does with light. I think if your roof wasn't in the way, you would see a ton better results cause you have a ton of light otherwise (well at least compared to mine LOL).

I doubt it makes a difference, but I don't copy to clipboard. I move the crop pointers and export to a file - I have noticed if I try to name the file at that point it gets wonky, so I let it save to the file name it wants and then change it later.
 
Yeah it is amazing how well the camera does with light. I think if your roof wasn't in the way, you would see a ton better results cause you have a ton of light otherwise (well at least compared to mine LOL).

I doubt it makes a difference, but I don't copy to clipboard. I move the crop pointers and export to a file - I have noticed if I try to name the file at that point it gets wonky, so I let it save to the file name it wants and then change it later.
I'm pretty sure that's what I have done in the past and not used the clipboard button. I think I panicked and did that because I thought it was wanting to export the whole 30min clip. I will try it again in a bit and see if it comes out as it should.
 
@Ri22o thanks for sharing the color night videos from the 5442. I know some people don't like a lot of lighting at night, but based on the neighbor's lighting across the street, it almost looks like with enough front lighting that 5442 could produce a decent image. Do you have more lighting available in the front?
BTW have you tried testing the mic on your 4k? My 4k-T has some crazy overpowering digital sounding noise in it.

@DanDenver I'm using HomeSeer as well. A long time ago I had it integrated with some Foscam and older Hikvision cameras to disable motion alerts when I was home. I see where other people have been trying to do automation for the same purpose with the Dahua cameras and newer HikVision cameras, but haven't really had any luck since especially Dahua is super clandestine about their API, and Hikvision's documentation is poor. I was hoping to be able to enable and disable Dahua IVS rules via API. It looks like I might have better success with HikVision. Or, I might just revert to ReoLink cams as their API guide is amazing and shows how I can easily enable/disable push notifications, email, etc. until Dahua or Hikvision sort all the issues out with their cameras.

I purchased Blue Iris many years ago, but need something energy efficient to run it on. It sounds like I'd have better luck doing what I want to do with HomeSeer and BlueIris, and integrating the cams into BlueIris?
 
Last edited:
@Ri22o thanks for sharing the color night videos from the 5442. I know some people don't like a lot of lighting at night, but based on the neighbor's lighting across the street, it almost looks like with enough front lighting that 5442 could produce a decent image. Do you have more lighting available in the front?
BTW have you tried testing the mic on your 4k? My 4k-T has some crazy overpowering digital sounding noise in it.

@DanDenver I'm using HomeSeer as well. A long time ago I had it integrated with some Foscam and older Hikvision cameras to disable motion alerts when I was home. I see where other people have been trying to do automation for the same purpose with the Dahua cameras and newer HikVision cameras, but haven't really had any luck since especially Dahua is super clandestine about their API, and Hikvision's documentation is poor. I was hoping to be able to enable and disable Dahua IVS rules via API. It looks like I might have better success with HikVision.

I purchased Blue Iris many years ago, but need something energy efficient to run it on. It sounds like I'd have better luck doing what I want to do with HomeSeer and BlueIris, and integrating the cams into BlueIris?
If all of the coach lights were white, like they typically are, and not orange/purple there would be more light. I will be able to get more light once everyone changes their lights, but I am not sure that would be enough.

I have not tested the speaker or mic.
 
@Ri22o I would love to know how the 5442 would do with more lighting. Let me know if you get a chance to test the mic.

@DanDenver
For some reason it was next to impossible to dig up much in the line of useful API info for controlling Dahua cameras with HomeSeer or other automation systems, but I was finally able to dig up some info, do some testing, and put together some examples to share:

 
When I first got BI with AI up and running I was so happy, alerts were actionable, meaning that when I got an alert it was because a human was exiting the sidewalk and approaching the house. Same for the back yard, if they exit the open space and approach the house, I would get an alert.

But then I would also get alerts because I opened the garage door to run an errand in my car. Or worse, I would just step out on my front porch and get an alert. Same for the backyard. So for a few weeks I would manually change BI to a profile that would not alert me under those conditions. Annoying, but BI was still new to me and I put up with it.

Then it occurred to me that I could tie BI into HS and all the sensors I already had around the house. Once having done that, BI works flawlessly! I get alerts while watching TV, running errands, while on vacation, etc. But if I step out onto the front porch, I no longer get alerts. Same for backyard.
Between me and my wife leaving/arriving entering/exiting I was getting so many false alerts. But with HS managing profiles automatically for me, if I get an alert now, I pay attention and it has made all the difference.

@IAmWatchingYou! Excellent post on Dahua camera calls. I call BI itself and effect change on the cameras via it's API rather than call the cameras directly. But I don't use IVS. I only use the AI in BI.

For example, when my backyard gate is opened, I direct my basement PTZ to zoom/aim to the gate for a short while. So in HS I call BI directly (using a CMD script) with this command:
curl -G h ttp:/192.168.1.45:81/admin --data-urlencode "camera=%1&preset=%2&user=xxxx&pw=xxxx"

A different example is profile management within BI:

If HS should change the BI profile I use a VB script to update BI to that change. Here is a snippet of that:
***
UrlString = "h ttp:/192.168.1.45:81/admin?profile=" & BI.Value & "&user=xxxx&pw=xxxx"
hs.URLAction(UrlString , "GET", "", "")
***


Should BI change the profile, I update HS with this call:
192.168.1.41/JSON?request=controldevicebyvalue&ref=387&value=%P
 
@DanDenver I'm really surprised I could only dig up a few other posts where people are talking about automating whether or not we receive alerts like we're doing with HomeSeer. What I was doing previously was using Tasker for Android to create http widgets on my home screen for Home, Away, Night, etc. to enable and disable motion detection on my cameras. I had to look up the HomeSeer API to figure out the direct http get I would need to do in order to activate those HomeSeer events, which would in turn use the camera's API to turn on/off motion detection. I'd like to automate those based on geolocation, but not sure I trust the powers that be with all the information, and I also have a roommate that wouldn't be onboard with geolocation, so I've been meaning to setup Bluetooth device proximity sensing with HomeSeer.

As I think through what I want to do now, I seem to be facing an issue that maybe you could shed some light on. I saw at least 1 other post where someone was wanting the same thing:

What I want to do for my outdoor cameras is have 2 regions with different actions.
Region 1: When a person or vehicle is detected on my property by the onboard AI, always record, but push notify only if armed by HomeSeer. (To cover these case when someone is home walking around outside, or on their way home.)
Region 2: When a person or vehicle is detected near my property (and could also be in Region 1). by the onboard AI, always record, but do trigger push notify.

From what I've read, this seems to not be possible with Blue Iris while using onboard camera AI. It seems that Blue Iris has no way of telling which AI/IVS rule, or generic motion detection got hit? So, it would be either get notified for everything including all cars and people nearby, or get notified for nothing? The only option I'd have with Blue Iris seems to be to configure the camera twice, one profile that would react to camera events that I could use for push notifications and to enable/disable push notifications, and a second profile that would use Blue Iris human/vehicle detection, which I read is pretty bad.

What I've done with the Dahua camera is to enable standard motion detection with both nearby and on-premises regions selected and enabled Smart Motion Detection. This is what I use for proximity recording. Then, I have an IVS intrusion rule that captures a human or vehicle actually on the property. The DMSS app lets you actually choose which type of events to subscribe to push notifications for. So, I choose to not subscribe to human and vehicle SMD events, but subscribe to IVS intrusion rule events. Now when I'm home, I can just have HomeSeer disable the IVS rule that I setup to detect when a person or vehicle enters the property, and voila, I still have recording, but no push notifications or email notifications.

From doing some reading, it also seems that what I'm doing with the Dahua cameras wouldn't even be possible with the Hikvision cameras because the HikConnect app forces you to subscribe to push notifications for All Events at once.
 
^Or it is the settings not dialed in yet or a problematic field of view. In my situation and field of view the motion blur is less with the 4K-T than the 5442.
 
  • Like
Reactions: CCTVCam
Region 1: When a person or vehicle is detected on my property by the onboard AI, always record, but push notify only if armed by HomeSeer
This does not make sense to me, "armed by HS"? HS can modify BI and BI can update HS as to its current state, but HS does not "arm" BI. However I may be completely missing your question.
For example, you can have HS tell BI to not send alerts during certain times of the day or based on the state of a given sensor.
I do this myself. If the back door is open, that tells HS to tell BI to not send me any alerts. Once the door is closed, then HS tells BI that it is ok to send alerts. Of course no matter what, the camera records all movement. Just the alerts are defeated.

From what I've read, this seems to not be possible with Blue Iris while using onboard camera AI. It seems that Blue Iris has no way of telling which AI/IVS rule, or generic motion detection got hit?
Why not use the zones in BI? You can easily handle the scenario above with zones and clones in BI.

and a second profile that would use Blue Iris human/vehicle detection, which I read is pretty bad.
This is an extreme position to take on the AI in BI. Many users are successfully using the AI in BI (including me) and have no issues. This is definitely a preference and an area where personal experience and bias plays a role. The AI in BI and the AI in cameras are all improving constantly and I would not direct anyone to one or the other. It is just about whatever makes you happy.
Note however, that when you get a lot of cameras managing them all individually for triggers/ivs/ai can introduce notable overhead. Also each camera needs to be updated in order to take advantage of any software improvements - which are inevitable in the field of AI.
I use BI AI as I have almost 20 cameras. I update Deepstack (or SenseAI in the future) and all the cameras benefit.
Note that changes in AI are coming fast and frequent. Just 16 months ago Deepstack was the preferred AI implementation and now everyone is flocking to SenseAI. I can't afford to replace all my cameras every time there is a significant leap forward in AI algorithms. But your budget may vary from mine.

it also seems that what I'm doing with the Dahua cameras wouldn't even be possible with the Hikvision cameras because the HikConnect app forces you to subscribe to push notifications
You appear to have a camera centric implementation, meaning that you rely on the cameras internal capabilities which as you have pointed out in this quote can lead to some minor limitations. I rely fully on BI and as such I can use any brand camera and get the same results (capabilities) from each camera

Now when I'm home, I can just have HomeSeer disable the IVS rule that I setup to detect when a person or vehicle enters the property, and voila, I still have recording, but no push notifications or email notifications.
I think I am missing the sentiment of this statement possibly. However it seems overly complicated to turn off the IVS rule just so that you don't get alerts. This can be done rather simply in BI. But if you choose to continue with this logic being in the camera, then maybe this is the best way for this use case.

Just some overall thoughts:
You bring up a good point about IVS versus AI (with person detection). I have to much wildlife to use IVS. I don't like getting false alerts at 2am because a racoon walks around my back porch. Visits from small wildlife on my property happens several times every week at a minimum.
So I only use AI. It is very effective/reliable. Roughly once every 2 or 3 months it will alert me at some late hour at night thinking that a cat or a coyote is a human. But that is not too many false alerts per year and I am ok with that.
I have read several threads in this forum and elsewhere wherein people strongly prefer IVS - the end goal, to get actionable alerts, is all that matters and luckily we have the options available to us to achieve this.

My first system was an NVR and I had IVS rules and regions. Can't count how many false alerts I got from that. So much time spent looking at clips to see what happened. Always just an errant bunny or shadow movement from the wind. After a while you get numb to all the alerts coming your way. With my current setup, if I get an alert, it is something that I immediately react to and respond.
 
  • Like
Reactions: jrbeddow
When I said "armed" I meant in the sense that I will have scenes in HomeSeer that will set my house in "armed" or "disarmed" security mode, and either have push notifications turned on in armed mode or turned off disarmed mode.

Thanks for the info about Deepstack. I guess I will just have to test it out to compare. What type of horsepower do you have BI running on that you can have Deepstack running for 20 cameras? I'm not sure how true this is, but one thing that I read people saying is that onboard camera AI detection is less resource intensive since it doesn't have to decode to do analytics. On top of that, if you can do analytics on the cameras, you wouldn't have to worry about CPU cycles on an NVR as much if you can just worry about the camera for triggering events. One big drawback with cameras triggering events in BI is it seems you can't differentiate between different rules that triggers an event.

Fortunately I don't have much in the line of trees on my property, but I do have a lot of wildlife. The most important thing for me in a camera and NVR system is going to be accurate push notifications, that can be disabled/enabled via automation, for anyone on my property, even more so that video quality. I am away for months at a time and no one should be on my property. Sadly both Hikvision and Dahua fail with audio.

That is a good point regarding updates - Deepstack is going to be receiving updates to it long after firmware upgrades end for cameras, and especially the AI in cameras.
 
What type of horsepower do you have BI running on that you can have Deepstack running for 20 cameras?
There are many threads on this subject and I have not researched this. I pay little attention to budget when it comes to my security systems as I will probably only need them once or twice in my life, but they damn well better work on those occasions. So I purchased a high end PC. Deepstack does not require ’massive horsepower’. It is widely adopted and utilized on a variety of PC hardware.
Your basic options for a PC setup are to run it on your CPU or your GPU. SenseAI is the same and my understanding is even less resource intensive than DeepStack.

onboard camera AI detection is less resource intensive
I have read this also and I am sure it is true. But with the number of cameras I run, the overhead of individually configuring and updating each camera is not attractive. My primary concern is that one camera will be effective at catching a ‘person’ and another will not. And with my luck, it will be the camera that should have triggered but did not. So I sleep better at night knowing that all the cameras have the same AI algorithm. If I have tested it on a few cameras, then all the remaining perform the same (by and large).
Sadly, I don’t have the time to test each individual camera after every firmware update.
Actually, with BI i simply do not update any firmware in any camera as it is not a required activity.

One big drawback with cameras triggering events in BI is it seems you can't differentiate between different rules that triggers an event
Again, a possibility is to only use BI for motion/triggers/rules/etc. It is effective and there are many who approach this challenge in this fashion. With this approach, you will always know what triggered what and how.
I would repeat my statement above, the possibility of one or more cameras not being properly setup exists when all the configuration is within each camera. You have to vet out each camera thoroughly (walking around testing: day motion/night motion/etc) in order to have confidence that it is responding as desired.
Not trying to sell BI - there are plenty of reasons to go with camera AI and triggers, it is a legit approach, just explaining some of the reasons I avoid a camera centric install.

Deepstack is going to be receiving updates to it long after firmware upgrades end for cameras
This is exactly my point. DeepStack is already no longer the favored solution for PC based AI. Who cares if it gets more updates? Any new install should in all probability be using SenseAI. DeepStack barely made it 24 months. Actually some were moving to SenseAI (BETA) around month 12. I will be moving to it this spring. Won’t cost me a dime and I won’t have to update each and every camera.
It won’t be long before the amount of memory shipped in current cameras won’t support the next major version release in camera embedded AI.
I think camera AI is excellent. But 6 years from now it will be a whole new playing field as AI is one of the most underdeveloped technologies around and companies are investing heavily in it.

Random thoughts:
You keep mentioning NVR. All my comments are for a BI install.

Any and all capabilities you may have available to you via any camera will be limited/enhanced by an NVR, so choose that unit well.
My NVR experience only found limitations.
 
There are pros and cons to every system and layout.

All we can do is share with you our experiences and then you have to decide what is best for you. And part of that involves testing a camera on your property with your field of view, which I know you are, but it still seems like you are trying to do too much with one camera.

Having one camera see a field of view and trying to do too much with one camera is much different than having every area covered by more than one camera. Too many people think one camera can be the be all/see all and they are the ones that are disappointed when something happens. We need more than one or four cameras, each one selected based on a primary goal.

In my situation, I have the field of view very well defined (not trying to do too much with one field of view) and the AI in the camera simply works as intended, so I do not care about any camera firmware updates to AI because why fix what ain't broke! I beta tested a newer AI firmware for a camera once and found the AI improvements to be worse than what I was running on the camera previously. Sometimes chasing "perfection" is a lost cause if the system already works for your needs.

However, I still also use Deepstack with BI on some of my cameras and as compliment to the camera AI for a few of the cameras. It flat out works for my situation so until a new BI update includes something I need that would force me to make the switch to SenseAI, I am staying put on my stable version of BI that absolutely rocks it for my situation.

I like redundancy as well, so in the event the BI AI konks out (which we have threads where DS or SenseAI has done that for some and they needed a restart to get it going again), I at least also have cameras that don't rely on that so that I still get an alert due to the camera AI.

In general using the resources of a camera instead of BI keeps down the CPU%, but that is so dependent on one's setup as well, not only the computer but the motion settings. Some people have BI motion triggered for EVERY tiny movement and sends it to AI for review. That can be resource heavy with enough cameras and the right situation (snow or rain constantly triggering and sending photos for AI).

Keep in mind most of the cameras have an audio input wire and you can get a cheap external mic to place next to the camera that will blow away any internal camera mic...
 
Last edited:
If
@DanDenver I'm really surprised I could only dig up a few other posts where people are talking about automating whether or not we receive alerts like we're doing with HomeSeer. What I was doing previously was using Tasker for Android to create http widgets on my home screen for Home, Away, Night, etc. to enable and disable motion detection on my cameras. I had to look up the HomeSeer API to figure out the direct http get I would need to do in order to activate those HomeSeer events, which would in turn use the camera's API to turn on/off motion detection. I'd like to automate those based on geolocation, but not sure I trust the powers that be with all the information, and I also have a roommate that wouldn't be onboard with geolocation, so I've been meaning to setup Bluetooth device proximity sensing with HomeSeer.

As I think through what I want to do now, I seem to be facing an issue that maybe you could shed some light on. I saw at least 1 other post where someone was wanting the same thing:

What I want to do for my outdoor cameras is have 2 regions with different actions.
Region 1: When a person or vehicle is detected on my property by the onboard AI, always record, but push notify only if armed by HomeSeer. (To cover these case when someone is home walking around outside, or on their way home.)
Region 2: When a person or vehicle is detected near my property (and could also be in Region 1). by the onboard AI, always record, but do trigger push notify.

From what I've read, this seems to not be possible with Blue Iris while using onboard camera AI. It seems that Blue Iris has no way of telling which AI/IVS rule, or generic motion detection got hit? So, it would be either get notified for everything including all cars and people nearby, or get notified for nothing? The only option I'd have with Blue Iris seems to be to configure the camera twice, one profile that would react to camera events that I could use for push notifications and to enable/disable push notifications, and a second profile that would use Blue Iris human/vehicle detection, which I read is pretty bad.

What I've done with the Dahua camera is to enable standard motion detection with both nearby and on-premises regions selected and enabled Smart Motion Detection. This is what I use for proximity recording. Then, I have an IVS intrusion rule that captures a human or vehicle actually on the property. The DMSS app lets you actually choose which type of events to subscribe to push notifications for. So, I choose to not subscribe to human and vehicle SMD events, but subscribe to IVS intrusion rule events. Now when I'm home, I can just have HomeSeer disable the IVS rule that I setup to detect when a person or vehicle enters the property, and voila, I still have recording, but no push notifications or email notifications.

From doing some reading, it also seems that what I'm doing with the Dahua cameras wouldn't even be possible with the Hikvision cameras because the HikConnect app forces you to subscribe to push notifications for All Events at once.
If using BI couldn’t you just use the blue Iris app and turn on geofencing?
 
Geofencing isn't accurate enough to turn off alerts when you're around the house mowing the lawn and doing other work. This would also entail all household members and/or employees of the business being OK with having the location of their mobile devices tracked. HomeSeer is capable of bluetooth proximity, but it's still subject to issues and all members of the residence to participate.
 
  • Like
Reactions: DanDenver
Geofencing isn't accurate enough to turn off alerts when you're around the house mowing the lawn and doing other work. This would also entail all household members and/or employees of the business being OK with having the location of their mobile devices tracked. HomeSeer is capable of bluetooth proximity, but it's still subject to issues and all members of the residence to participate.
I didn't realize that. That sounds pretty cool.