Review-OEM 4mp AI Cam IPC-T5442TM-AS Starlight+
- Thread starter looney2ns
- Start date
Hi @looney2ns, thanks for your efforts in this review.
From your testing, do you have any comments on how well the AI (people, vehicle detection) worked, especially at night?
The ability to discriminate between people and cats/dogs would mean I could stop all the false PIR triggering of external lights due to animals.
From your testing, do you have any comments on how well the AI (people, vehicle detection) worked, especially at night?
The ability to discriminate between people and cats/dogs would mean I could stop all the false PIR triggering of external lights due to animals.
CCTVCam
Known around here
- Sep 25, 2017
- 2,815
- 3,683
Dr. Looney2S: A or B...1 or 2
Me: Ugh....
Mark
That to me Mark just screams insufficient bandwidth. Looking at the 4mp screenshot, the image is over sharpened and full of artefacts - a particular area that stands out in comparison between the two is the road. All of what I'm seeing points to over compression. That might not be BI but it could be Dahua's in camera compression rate. I rather suspect looking at it they're trying to compress too much info into too little bandwidth.
Here are excerpts from both photos at 100% (need to be viewed also at 100% to see the difference, so you might need to expand or follow the link).
Whilst the eye test board is sharper on the 4mp, the camera is using noticeably increased image sharpening settings compared to the 2mp in my opinion, which explains the difference.
There are several areas showing artefacts through over compression, the 2 areas giving the game away are the driveway and the road. Again, this might not be noticeable if you view the pictures at less than 100% true size.
With the driveway, on the 2mp, you can see the flecks of stone that make up the paving slabs. Compare that to the 4mp version, and you find that detail is lost and instead totally replaced with a compression artefact filled interpretation of what is on the left. It's quite obvious the info has been discarded and the codec has tried to interpret what texture to replace it with when decompressing the photo. The same can be said of the road. It's not perfect in the 2mp picture either and some compression artefacts are probably inevitable with CCTV data rates, however, the 4mp picture is noticeably worse and loses a lot more detail and suffers a lot more artefacts. There are other areas such as Mark's forehead especially around the eye area, and arms.
Direct link i case can't see it on the forum at 100% sample:
What really matters though is overall, eye test board aside, I find left Mark sharper than right Mark overall and easier to identify.
I think maybe Dahua need to look at their compression with a view to setting new data rates.
Dramus
Pulling my weight
Thank you, CCTVCam! I knew there was something bothering me about the 4mp vs. the 2mp images, but did not know how to put it into words--much less have a technical explanation for it. I think you've nailed it. This may explain why some of us look at the A/B images and are underwhelmed with the newer product's images.
I note @SecuritySeeker made similar comments wrt to some low-light images a page back.
@looney2ns: I don't recall, or recall if you even mentioned it, but both cameras were using the same compression, yes? (Not that I expect you'd make such a n00b mistake, but, with @CCTVCam's remarks, the question must be asked.)
I note @SecuritySeeker made similar comments wrt to some low-light images a page back.
@looney2ns: I don't recall, or recall if you even mentioned it, but both cameras were using the same compression, yes? (Not that I expect you'd make such a n00b mistake, but, with @CCTVCam's remarks, the question must be asked.)
dahua guy told me Smart IR is same as the current IR Auto we use on the new 4.0AI camera.
I found it to be surprisingly good. I don't get a lot of traffic after dark, but it did well the the few times I was able to try it out. Since, there is new firmware, I will test that more in depth, as I now have a little more free time. I have plenty of moving tree shadows to give it a work out during the day.
Yes, they were using the same compression, I also did a trial using different compression types to see if that was the cause, but it didn't cahnge anything.
@EMPIRETECANDY has recently obtained new firmware for the cam, and It has fixed some original issue's, such as "auto" Ir not working. I've not had time to do a new side by side, but will very soon.
The stills with the eye chart were taken directly from the sd cards on the cam.
Last edited:
dahua guy told me Smart IR is same as the current IR Auto we use on the new 4.0AI camera.
It is now working that way with the new firmware.
bigredfish
Known around here
One thing I appreciate on my Dahua CVI cams is that the SmartIR has a sensitivity setting 1-10 to adjust how "Smart" it reacts to objects close by.... you would think they could implement that in the IP cams..
redfive
Pulling my weight
- Apr 13, 2016
- 514
- 207
Andy, I'm not comparing the 4.0-AI VS eg. the 5231R-ZE, I mean that the SmartIR, even on the ultra series, such as the HFW8232E-ZE, or the HFW81230E-Z, IMHO, is not even close to the Smart IR of the Hikvision DS-2CD2042WD-I, and I'm not an Hikvision fan, since I bought the first dahua camera (from you), I've only bought dahua...dahua guy told me Smart IR is same as the current IR Auto we use on the new 4.0AI camera.
CCTVCam
Known around here
- Sep 25, 2017
- 2,815
- 3,683
The trouble with the same compression is it's trying to fit twice as much picture detail information into the same amount of bandwidth ie 4 million pixels worth instead of 2 million pixels worth. So technically, on paper it's not the same compression, theoretically you've doubled it (or by doing so, halved the quality) ie so whilst the available space has remained the same, the amount of information you have to fit into it has doubled. In turn, that means more information has to discarded and recreated as a texture reference (with mistakes) to get it to fit.
Also, if you're just changing compression in the recorder, it's not going to make 1 bit of difference if the camera has already compressed it and discarded information, as you can never get that back.
In reality, is not strictly true that you've doubled the compression, as the amount of info recorded and not referenced, key frames aside, is dependent on the amount of detail in the shot plus the amount of change between frames. A shot with a lot of texture such as the paving slabs, road, grass blades or leaves will potentially contain more information and thus require more compressing although some of that will be alleviated if there is no movement. In a low detail shot with no movement, there is a lot less to record rather than reference so the difference will be much less. The detail factor really shows in the shot above. At 2mp, the compression is able to keep a lot of the fussy detail. At 4mp, the need to cram 2m extra pixels in there means it's having to discard a lot more of the areas of fussy similar detail and instead replace whole areas with interpreted detail, as the amount of detail is so high, it won't fit. The issue is guessed (interpreted) similar textures aren't always correct. Hence the artefacts and soft look in those areas. The answer here is to increase the bit rate to compensate. Unfortunately, there are 2 places where bit rate is fixed, in the camera and in the recorder / recording software. In the camera it's fixed by the manufacturer and in the recorder / software, there may be some adjustment, but it's very limited. I have commented before that I thought 8mbs was somewhat low for 1080P, but this is CCTV and not Hollywood! For potentially double the info, it currently looks very restrictive. That said the recorder / software is always going to be constrained by the camera stream. There's little point recording a stream at say 14mbs in the recorder if the camera is streaming it at 6mbs. Yes there will be less compression on the final recording, but the damage will already have been done by the camera.
I believe these cameras are on Amazon but they say they are by Loryta . What is Loryta ?
Are there differences between Andy directly, Aliexpress, and Amazon? Looks like they all sold by Andy.
Are there differences between Andy directly, Aliexpress, and Amazon? Looks like they all sold by Andy.
Loryta is my registered brand in US , empiretech still under registering. All will be by us. All are dahua original International updated cams.
How does daylight video compare to a 8mp camera, like IPC-HFW1831E?
Will I miss a lot of details?
Will I miss a lot of details?
giomania
IPCT Contributor
- Jun 1, 2017
- 784
- 540
Remember when I said you should have gone with “AndyCam”?
Sent from my iPhone using Tapatalk Pro
For what it's worth, here are some more stills at different bit rates.
Have at it.
2mp H264 4096br
6144br h264
8192br h264
2mp H264H
4096
2mp H264H 6144
2mp H264H 8192
4mpAI h264 4096
4mpAI h264 6144
4mpAI h264 8192
4mpAI h264h 4096
4mpAI h264h 6144
4mpAI h264h 8192
Have at it.
2mp H264 4096br
6144br h264
8192br h264
2mp H264H
4096
2mp H264H 6144
2mp H264H 8192
4mpAI h264 4096
4mpAI h264 6144
4mpAI h264 8192
4mpAI h264h 4096
4mpAI h264h 6144
4mpAI h264h 8192
Attachments
VisionxOrb
Getting the hang of it
(I've been lurking so far because I'm in the market for some more cams. Haven't been paying attention for about a year and was wondering if there's a good 5231 replacement)
With 8mbit h265 the compression artefacts are mostly gone and I'm actually pretty impressed with this cam. This, combined with the nighttime performance almost makes me get this over a 5231... I think there's another thread with a head-to-head shootout. What do you guys think (assuming 6mm is enough zoom)?
With 8mbit h265 the compression artefacts are mostly gone and I'm actually pretty impressed with this cam. This, combined with the nighttime performance almost makes me get this over a 5231... I think there's another thread with a head-to-head shootout. What do you guys think (assuming 6mm is enough zoom)?
