Mini-Review EmpireTech IPC-Color4M-T

wittaj

IPCT Contributor
Apr 28, 2019
27,511
53,461
USA
Based on this thread about the lower end 2449-PRO and the better 3449-PRO, many of us wanted to see how it would perform in our situations instead of ideal situations in a marketing video.

The EmpireTech 4M-T camera boasts the following:
  • 4MP on the ideal 1/1.8" sensor
  • Two-way talk
  • Infrared and white light
  • Red/blue siren lights
  • Audio alarm capabilities
The full capabilities and purchase options can be found on @EMPIRETECANDY website and Amazon store. As of right now, it is only available in the 2.8mm version.

My first impressions are that the image is comparable to the original 5442 fixed lens camera, but it benefits from a better sensor (but not the S3) along with stronger infrared.

I have found this two-way talk to be the fastest of any of the other cameras I have with two-way talk. It didn't suffer from the lag that many of us are used to.

The speaker is fairly loud for a turret. Not as loud as the 4K/X bullet, but louder than the 180T.

The red/white flashing light is way brighter than the previous TIOC cameras. The first generation you could only really see the flashing if you looked up at the camera. This one illuminates the area around the camera and the light can be seen reflecting off the neighboring house. The previous TIOC had no chance at doing that. In the front of the house, one may think it is police heading that way.

I have stated this before, but the more I play with the new GUI on my other cameras that have the new GUI (4K-T, mini-PTZ, T180) the more I start to like it. It has way more options and flexibility. It still has some quirks as well LOL.

This model has AcuPick like the other models with the newer GUI, but as mentioned, you need an NVR capable of using this feature for it to work.

I won't bore you with the GUI screenshots as they are getting fairly standard across the board now with the new GUI.

Here are some first impression night pics while I work on getting some additional comparisons.

As usual, I am using my two-faced neighbor as the model LOL. This time it is his kid doing a light jog/fast walk for these images. Settings are 1/120 shutter, 0-50 gain, 40 NR. Other settings adjusted a few off of default 50.

The first image is just the infrared - you will notice a tree shadow in the background that is from another camera, but the house is blocking it from providing any additional lighting to the person at these locations:

TIOC IR2.jpg


The next is the white LED but on auto day/night so it stayed in B/W. This allows you to see some features on the clothing that was lost with infrared:

TIOC LED BW.jpg


This is forcing the camera into color with the white LED on. While the B/W can see a little further out, I like this image better (although some may not like the white light):

TIOC Color.jpg

And then one of the things we talk about is the exposure can get messed up with the flashing light a bit, but it also shows how much stronger it is than previous TIOC:

TIOC flashing.jpg


I am still playing with it and will try a few more things, but a quick summary is this is the first TIOC that I think we can recommend. Previous TIOCs suffered from less than ideal MP/sensor ratios and this is the first one I recall that is on the ideal MP/sensor ratio.

This one is giving us the image quality of the 5442, but adds a speaker, which many people like.

The ability to use white light or infrared is a huge selling point.

The additional TIOC features are bonuses if you want to use them.

The price-point is basically the same as the 5442, but with these additional features. I am a big fan of at least one camera with a speaker on each side of the house to hopefully send out alarms and this one provides the image quality that was lacking on many other cameras with that feature.
 
Last edited:
As an Amazon Associate IPCamTalk earns from qualifying purchases.
I have installed & used 8 TIOC-PRO cameras in various installations for about a month. Since then I planned to do a review, but according to my information these models still do not have the final version of the firmware released, using the full capabilities of AI-ISP (WizColor).

Since @wittaj started the thread, I will write a few sentences about this camera from my observations. A very important note: my comment about those cameras are based on latest NOT OFFICIAL (BETA) firmware version, which can only be downloaded from the Dahua Poland server. The factory firmware has other, worse characteristics.

The 3449-PRO (TIOC-PRO) camera is an evolution where Dahua took the regular TIOC camera on a 1/2.8" sensor, changed the sensor to a larger 1/1.8" one, put brighter hybrid lights (dual white + IR in the same LED) and introduced the next generation of image processing.
Processing which is a next generation extension of what we saw on Full Color cams (Dahua now calls it WizColor). According to Dahua, these cameras are the first to use AI-based processing (AI-ISP). Wiz-Color page: undefined - Dahua International

These cameras use a different sensor than the 5 and 7 series cameras (5442, 5449, 7442 which are OmniVision based). TIOC-PRO are available in 4/6/8Mpx resolutions and all models have a WDR of 120db, which means that the sensor used here is identical to the HikVision ColorVu 2.0 series cameras (SmartSens sensor).

From my observations at short shutter speeds (1/250 - 1/1000), in night conditions with 3D reduction turned off - this new sensor has completely different characteristics of generated noise. The sensor in 5442 (OmniVision) is sensitive but can also generate very characteristic bright noise. The noise generated in the TIOC-PRO model is different - is more dark noise similar to Sony's Starvis sensors.

Comparing 5442 and TIOC-PRO in difficult conditions with 3D noise reduction turned off - for me TIOC-PRO has about 2x advantage in input image sensibility compared to 5442-S3.

Image processing is added to this. AI-ISP in TIOC-PRO uses 2 technologies, in a much more advanced version than previous camera models. The first is HDR (high dynamic range), known from mobile phones and sports cameras and being a variation of single-frame WDR (with values <45) from CCTV cameras. HDR causes very dynamic remapping of the image gamma curve (image brightness) and the LUT color table. Those dynamic remappings are very local to different regions of image (so dark part of image have different gamma/color curve as bright ones).

HDR is always turned on in these cameras and in fact the image in standard mode and single-frame WDR with WDR values at around 30-40 is practically no different. Using HDR generates very bright/readable and also very colorful images even in very dark image conditions. This image doesn't look at all like old FULL WDR from existing Dahua cams (dull and not natural).

The second technology is high noise reduction based on averaging neighboring video frames (3D noise reduction). This technique, combined with HDR, generates fully colorful images with practically zero lighting. The downside is of course movement - which immediately reveals that TIOC-PRO cameras (and AI-ISP processing) cheat a lot - moving people are noisy and not sharp.

Interesting about AI-ISP processing is that many image settings on those cams works different or don't works at all (sharpness, gamma). More interesting there is a new image settings section called AFSA (anti flicker scene adaptation) which can create even better/brighter image at very dark night conditions - but this settings disables (mostly by ignoring) almost all image settings (even shutter or backlight settings). Another discovery was that enabling full 2-shot WDR (with values>45) disables AI-ISP processing and enables classic Dahua full WDR algorithm.

In practical applications, where the 5442-S3 in forced color mode generates a very dark, very illegible image, the 3449-PRO generates a beautiful color static image, which can be very misleading as to the camera's capabilities. Of course, movement (people, animals, trees in wind, cars) immediately reveal the level of deception. In poor lighting conditions, they are noisy and partially transparent (ghosting created by 3D noise reduction).

Of course, I remind you that in such conditions the 5442-S3 cannot handle anything in color mode (both static image and movement) - so here the TIOC-PRO has an advantage.

For me, 2 things are important:
  • even in difficult lighting conditions, where people / movement don't look good, motion detection (SMD) and IVS work very well in forced full color mode - so you can leave these cameras in color mode at night to detect movement.

  • TIOC cameras in the standard setting at night turn on white light when they detect human (or car) movement. And this white light works very well to illuminate a very dark place and increase movement / human image quality / visibility.
Of course, in addition to white light, armed (like an alarm system) TIOC-PRO cameras can generate (by IVS activation) an audio-visual alarm, where a loud siren signal (or a selected/uploaded audio message) is also combined with flashing clear police lights (blue-red). This can works as a deterrent mechanism that works very well against intruders and strange individuals.

TIOC-PRO is first camera for me, which can works in full color even at very bad nigh conditions. It can detect people and activate white light to have good quality image. In case of 5442, we have to use IR for that conditions (black-white image). Of course we can use also IR in TIOC-PRO, this camera have big flexibility of illuminator settings - we can choose IR or white light, we can also enable white light activation on movement (or switching from IR to white on movement).

The biggest drawback of TIOC-PRO cameras:
  • they are only fixed lens options (2.8 and 3.6mm). There is no varifocal options (like new 5449-ZE Full color varifocal constant F1.0 model).

  • AI processing is limited in comparison to 5442-S3. AcuPick is only 1.0 version, there is no Video Meta Data (so no people/car attributes/snapshots), there is no support for animals, IVS self-learning or long-mode.

  • they cheat, can generate beautiful image for bad night conditions, but any movement shows that this is a deception. Yes they have better bad light / night conditions capabilities that any 5442-S3, but You always must check / manually configure - how they work for movement.

Comment added later: for me, TIOC-PRO cameras, despite current limitations or problems, are a big preview of what future cameras from Dahua or HikVision will look like.

I won't be surprised at all that the next generation of 5442 will use more refined technologies used today in TIOC-PRO (AI-ISP processing using strong HDR and 3d NR, hybrid IR/white lighting with motion activation, maybe even a speaker/siren/alarm LEDs) but in a package & characteristic of the PRO line (better image sensor/lenses, large selection of variants and optics, mega-extended AI).
 
Last edited:
Based on this thread about the lower end 2449-PRO and the better 3449-PRO, many of us wanted to see how it would perform in our situations instead of ideal situations in a marketing video.

The EmpireTech 4M-T camera boasts the following:
  • 4MP on the ideal 1/1.8" sensor
  • Two-way talk
  • Infrared and white light
  • Red/blue siren lights
  • Audio alarm capabilities
The full capabilities and purchase options can be found on @EMPIRETECANDY website and Amazon store. As of right now, it is only available in the 2.8mm version.

My first impressions are that the image is comparable to the original 5442 fixed lens camera, but it benefits from a better sensor (but not the S3) along with stronger infrared.

I have found this two-way talk to be the fastest of any of the other cameras I have with two-way talk. It didn't suffer from the lag that many of us are used to.

The speaker is fairly loud for a turret. Not as loud as the 4K/X bullet, but louder than the 180T.

The red/white flashing light is way brighter than the previous TIOC cameras. The first generation you could only really see the flashing if you looked up at the camera. This one illuminates the area around the camera and the light can be seen reflecting off the neighboring house. The previous TIOC had no chance at doing that. In the front of the house, one may think it is police heading that way.

I have stated this before, but the more I play with the new GUI on my other cameras that have the new GUI (4K-T, mini-PTZ, T180) the more I start to like it. It has way more options and flexibility. It still has some quirks as well LOL.

This model has AcuPick like the other models with the newer GUI, but as mentioned, you need an NVR capable of using this feature for it to work.

I won't bore you with the GUI screenshots as they are getting fairly standard across the board now with the new GUI.

Here are some first impression night pics while I work on getting some additional comparisons.

As usual, I am using my two-faced neighbor as the model LOL. This time it is his kid doing a light jog/fast walk for these images. Settings are 1/120 shutter, 0-50 gain, 40 NR. Other settings adjusted a few off of default 50.

The first image is just the infrared - you will notice a tree shadow in the background that is from another camera, but the house is blocking it from providing any additional lighting to the person at these locations:

View attachment 206998


The next is the white LED but on auto day/night so it stayed in B/W. This allows you to see some features on the clothing that was lost with infrared:

View attachment 206999


This is forcing the camera into color with the white LED on. While the B/W can see a little further out, I like this image better (although some may not like the white light):

View attachment 207000

And then one of the things we talk about is the exposure can get messed up with the flashing light a bit, but it also shows how much stronger it is than previous TIOC:

View attachment 207001


I am still playing with it and will try a few more things, but a quick summary is this is the first TIOC that I think we can recommend. Previous TIOCs suffered from less than ideal MP/sensor ratios and this is the first one I recall that is on the ideal MP/sensor ratio.

This one is giving us the image quality of the 5442, but adds a speaker, which many people like.

The ability to use white light or infrared is a huge selling point.

The additional TIOC features are bonuses if you want to use them.

The price-point is basically the same as the 5442, but with these additional features. I am a big fan of at least one camera with a speaker on each side of the house to hopefully send out alarms and this one provides the image quality that was lacking on many other cameras with that feature.

@wittaj are you upgraded to this test firmware on TIOC-PRO, to one I sent you? Your pictures looks like factory (non AI-ISP) firmware...
 
Last edited:
As an Amazon Associate IPCamTalk earns from qualifying purchases.
Great write up!! The new firmware you talk about that unlocks the cameras full potential, do you know when this will be available and will it be through Empire Tech or Dahua? Thank you sir

When they finish it ;)
As I understands Dahua is not happy from movement (people, cars) performance at this camera models, so they try to correct this.
Learning a better AI models takes a long time (depending how big AI datacenter / AI servers Dahua have access) and this is repetitive process - sometimes new trained AI model is better sometimes is much worse..

For now You have working beta, to which link I gave in my long post..
 
When they finish it ;)
As I understands Dahua is not happy from movement (people, cars) performance at this camera models, so they try to correct this.
Learning a better AI models takes a long time (depending how big AI datacenter / AI servers Dahua have access) and this is repetitive process - sometimes new trained AI model is better sometimes is much worse..
Makes sense! Thank you for the quick reply sir!
 
@wittaj are you upgraded to this test firmware on TIOC-PRO, to one I sent you? Your pictures looks like factory (non AI-ISP) firmware...

Not yet. I wanted to test with the stock version first. Is that link above international or will the camera go to a foreign language LOL


Thanks for @wittaj sharing, the camera is working bit better compare to the old 1/2.8cmos.

Yes it is. I am glad to see they are investing on cams that can produce a quality image and provide TIOC capabilities!
 
Not yet. I wanted to test with the stock version first. Is that link above international or will the camera go to a foreign language LOL




Yes it is. I am glad to see they are investing on cams that can produce a quality image and provide TIOC capabilities!
lol.. this is normal International firmware… simply not published on Dahua website - but available to Dahua resellers on some internal Dahua server

Try it… a do also comparisons for different shutter times (1/30 to 1/250). You will be surprised how different this camera works. Or put any pictures from 5442 in color mode for comparison
 
Last edited:
I have to withdraw my recommendation to install the latest BETA version. I've only had it for a few days and only today had a chance to test it longer & properly. It's much brighter than the previous beta, but has a lot more static noise at short exposure times.

Below is my archive of all the firmware versions for TIOC-PRO that I managed to get:


The previous beta version - DH_IPC-HX3XXX-Goethe_MultiLang_PN_Stream3_V3.140.0000000.3.R.240905.bin worked the longest (almost a month) and best on my cameras. I recommend this one...
 
  • Like
Reactions: redfive
Compare to the Sep firmware, the Sep fw not support AI-SIP

1731397166087.png


1731396900946.png
 
Compare to the Sep firmware, the Sep fw not support AI-SIP

View attachment 207069


View attachment 207066

Yes, I know.. latest firmware is much bigger I behave totally different comparing to all previous one...

previous used classic 3D NR algorithm (plus strong HDR gamma/LUT remappings). I have many installations where one part of image have some light from the street and second one is totally in dark. I brighten up those parts using exposure/gamma settings and classic 3D NR usually removed 90% of noise. That remaining 10% was visible as small changes every 2 seconds (I-frame) but totally they didn't bother.

static color image looked very good, when camera detected people extra white light allowed to capture nice and sharp pictures of people. TIOC-PRO worked for that half lighted half totally dark locations much better that 5442-S3 in any color settings.

Now I see that classic 3D NR was replaced by AI NR which creates totally bright & static image for that dark corners. There is no need to use higher exposure/gamma - AI-ISP is doing this alone. Everything is bright. But I see that AI NR invents what is in the picture, drawing textures that are not there for that totally dark corner at fast shutter time (1/100). It looks like some strange static noise (it not change at all). And this can't be corrected.

Dahua promised on the WizColor page that they would release this firmware for TIOC-PRO with AI-ISP in October. They didn't do it (this is private beta) - so I conclude that they themselves know that this AI still needs to be worked on and trained a lot.

From the Polish video demonstrating the comparison of 2449-PRO and 3449-PRO to 5449 Full Color and the old TIOC in total darkness (forest) you can see that 2449-PRO performs much better than 3449-PRO. Sensor/lenses are the same, so it looks like differences in SOC and firmware.

I'm not afraid of further development of TIOC-PRO. Hik-vision announced 3rd generation ColorVu a few weeks ago, based on the same sensors set and SOC with AI-ISP (specific features of sensors and SOCs matches with TIOC-PRO in 100%). ColorVu is their most important line of cameras, so I'm sure they will force the SOC supplier (I don't know if it's HiSilicon or someone else) to provide refined AI-ISP models.

As each AI model it simply needs lot of time, AI computing power and talented AI specialists.
 
Last edited:
Can anybody who has one of these Color-4M-T cams tell me for certain if it's possible to configure MJPEG encoding on any stream with a resolution of 1920x1080 or higher? If so, what is the highest bitrate allowed specifically for MJPEG?

I have been able to configure MJPEG at 1920x1080 at 20Mbits on my T5442T-ZE (presume it's an -S2, with the grey/yellow GUI), but only on substream #2!

None of my cams with the newer white/blue WebUI will do MJPEG at any resolution higher than 704x480 (D1).. :(

Thanks in advance..
 
Can anybody who has one of these Color-4M-T cams tell me for certain if it's possible to configure MJPEG encoding on any stream with a resolution of 1920x1080 or higher? If so, what is the highest bitrate allowed specifically for MJPEG?

I have been able to configure MJPEG at 1920x1080 at 20Mbits on my T5442T-ZE (presume it's an -S2, with the grey/yellow GUI), but only on substream #2!

None of my cams with the newer white/blue WebUI will do MJPEG at any resolution higher than 704x480 (D1).. :(

Thanks in advance..

Yes, on TIOC-PRO (Color-4M) You can do 1080p MJPEG on first sub stream...

Screenshot 2024-11-12 at 15.57.40.png