Dahua WizColor 5x59-PRO and SmartLight 5x59-IL new series

[steve1225]

The question isnt the FW - its the sensor

@bigredfish

We don't know that.. SmartSense sensors were used for many years in non AI-ISP (classic video processing) Full Color cams (HIK vuColor and some Dahua Full Color) and they have comparable night color performance to 5442...
For some reason they were used in Full Color cams - probably had better color night performance and worse IR one to OmniVision sensors...

All problems we see with TIOC-PRO and 5459-PRO looks like totally different processing using AI-ISP magic box - totally fucked up..

Different 2D/3D NR plus always-HDR, which can't be disabled. It brighten up an image a lot using strong gamma manipulation. Side effect is increased a lot image noise.

You can see what happens on 5442 at night on moving objects when you brighten up image a lot using brightness and gamma image settings on short shutter speed - the same effect.

 

[Arjun]

@bigredfish

We don't know that.. SmartSense sensors were used for many years in non AI-ISP (classic video processing) Full Color cams (HIK vuColor and some Dahua Full Color) and they have comparable night color performance to 5442...
For some reason they were used in Full Color cams - probably had better color night performance and worse IR one to OmniVision sensors...

All problems we see with TIOC-PRO and 5459-PRO looks like totally different processing using AI-ISP magic box - totally fucked up..

Different 2D/3D NR plus always-HDR, which can't be disabled. It brighten up an image a lot using strong gamma manipulation. Side effect is increased a lot image noise.

You can see what happens on 5442 at night on moving objects when you brighten up image a lot using brightness and gamma image settings on short shutter speed - the same effect.

@steve1225 I have to agree with @bigredfish that the SmartSense sensors are lacking quality versus the Omnivision sensor. The colors in the Color-4M-T (I own two of the fixed turret models and 1x Color-4M-TZ) are dull and washed out, can't really tell the difference in Vivid, Soft, Normal, BLC, HLC, and SSR. WDR is hit or miss.
The daytime quality of the T54PRO is decent, but at night it suffers when it switches to color when the LEDs turn on, and I am forced to keep it in B&W after sunset. What was the point of that when the 5442 was working perfectly.
It would not cost much more keep the 5442's in inventory; technically it is more expensive to maintain inventory of the AI models due to more appetite for DDR

 

[wittaj]

@steve1225 I have to agree with @bigredfish that the SmartSense sensors are lacking quality versus the Omnivision sensor. The colors in the Color-4M-T (I own two of the fixed turret models and 1x Color-4M-TZ) are dull and washed out, can't really tell the difference in Vivid, Soft, Normal, BLC, HLC, and SSR. WDR is hit or miss.
The daytime quality of the T54PRO is decent, but at night it suffers when it switches to color when the LEDs turn on, and I am forced to keep it in B&W after sunset. What was the point of that when the 5442 was working perfectly.
It would not cost much more keep the 5442's in inventory; technically it is more expensive to maintain inventory of the AI models due to more appetite for DDR

I suspect difficulty in getting the sensors are the limiting factor for the 5542.

The alternative is to come out with 5442-S4 with the crappy sensor and that will just tick more people off.

 

[alekk]

@Arjun : WRT to "dull/washed out" colors on the 4M. I don't have that, but on the 54PRO, it "boosts" the black ... so you don't have a full dynamic range - you can see this in the histogram. If you set Backlight->HLC=1 (heard this from the always smart @steve1225), and you get more dynamic range and the picture doesn't look as dull.

 

[Arjun]

@Arjun : WRT to "dull/washed out" colors on the 4M. I don't have that, but on the 54PRO, it "boosts" the black ... so you don't have a full dynamic range - you can see this in the histogram. If you set Backlight->HLC=1 (heard this from the always smart @steve1225), and you get more dynamic range and the picture doesn't look as dull.

I did notice that on HLC but depending on the weather - cloudy skies make the 4M colors duller versus my 5442 and 54PRO. My 54PRO does well in full dynamic range during the day. Its only the 4M in which the colors are washed out and many others on the forum confirmed the same observation

 

[Arjun]

I suspect difficulty in getting the sensors are the limiting factor for the 5542.

The alternative is to come out with 5442-S4 with the crappy sensor and that will just tick more people off.

They should just come out with a new model number approach at this time.

 

[wittaj]

They should just come out with a new model number approach at this time.

That was the PRO series LOL

 

[Arjun]

@Arjun : WRT to "dull/washed out" colors on the 4M. I don't have that, but on the 54PRO, it "boosts" the black ... so you don't have a full dynamic range - you can see this in the histogram. If you set Backlight->HLC=1 (heard this from the always smart @steve1225), and you get more dynamic range and the picture doesn't look as dull.

I also have to add that the Color-4M reminds of the old analog BNC camera days - the image quality is that bad - its appalling!

 

[Arjun]

Here is the vinyl siding of my neighbor's house (captured from a 5442-S3 ZE)




VERSUS the Color 4M (3.6mm Fixed Turret Model)



Image settings are the same between both models - can you spot the difference with respect to color saturation?

 

[Techie007L]

@Arjun That's because the IPC-Color4M-T cameras are trash, and EmpireTech is actively hiding bad reviews of them on their site (you really ought to leave another one, with multiple image comparisons like the above). Should've returned it and bought two of the vastly superior $95 IPC-Color4M-TL cameras instead! Just be aware that the IPC-Color4M-TL has a few quirks that need to be adjusted for to maximize performance:

• It uses a 1440p sensor despite having a 1520p encoding option, so be sure to run it at 1440p resolution to avoid blurriness and grid-like interpolation artifacts.
• For best image, start with Self-adaptive mode, Standard style, Saturation=55, Sharpness=25, Gamma=45, and 2D NR Level=15. Note that after setting these and exiting out of the camera UI, you will need to fully reboot the camera from Blue Iris due to the widespread firmware bug that reverts the 2D NR Level back to 50 upon closing the webpage (it'll reload your set value upon reboot). Note that if you fail to drop 2D NR Level, the camera will have poor image clarity; and if you fail to drop Sharpness, the camera will have an almost embossed image—the defaults really do it no justice!
• If you find the nighttime image too bright, I found it can be darkened by setting Illumination Overexposure Remover to Number Plate Priority. Note that this also causes the camera to flash sometimes, particularly around moving vehicles. While I don't care for the flashing, it does make it capture sharp plate images and choose a reasonable image brightness at night.

With these settings applied, I believe you will find the IPC-Color4M-TL will take pictures just as clear and vibrant as the 5442, with less noise and significantly better nighttime performance. It feels like the "little brother" to the IPC-Color4K-T and is now my "go to" camera unless visible illumination at night is undesirable. Too bad they don't come in a 6mm variety or varifocal!

 

[steve1225]

@steve1225 I have to agree with @bigredfish that the SmartSense sensors are lacking quality versus the Omnivision sensor. The colors in the Color-4M-T (I own two of the fixed turret models and 1x Color-4M-TZ) are dull and washed out, can't really tell the difference in Vivid, Soft, Normal, BLC, HLC, and SSR. WDR is hit or miss.
The daytime quality of the T54PRO is decent, but at night it suffers when it switches to color when the LEDs turn on, and I am forced to keep it in B&W after sunset. What was the point of that when the 5442 was working perfectly.
It would not cost much more keep the 5442's in inventory; technically it is more expensive to maintain inventory of the AI models due to more appetite for DDR

All options you mentioned are video processing options done in SOC (processor) - in mixed software & hardware solution. Not at image sensor level.

The only image settings which change image sensor behavior are: shutter time, gain (aka ISO) and full WDR (>= 45 or 50 depending of camera model).

Most people here don't understand that RAW data from image sensor and what we see as image/video in JPG or H264 - are totally two different things.

Especially at night. There are TONS of processing between.. And all are done by image signal processor (ISP) in SOC under of control of firmware.

this is why 5442-s3 is good... and those new AI-ISP models from Dahua or new 1/1.2" sensor cams from Ubiquity ARE NOT..

5442-s3 have very good image processing which process data from very good image sensor..

 

[Arjun]

All options you mentioned are video processing options done in SOC (processor) - in mixed software & hardware solution. Not at image sensor level.

The only image settings which change image sensor behavior are: shutter time, gain (aka ISO) and full WDR (>= 45 or 50 depending of camera model).

Most people here don't understand that RAW data from image sensor and what we see as image/video in JPG or H264 - are totally two different things.

Especially at night. There are TONS of processing between.. And all are done by image signal processor (ISP) in SOC under of control of firmware.

this is why 5442-s3 is good... and those new AI-ISP models from Dahua or new 1/1.2" sensor cams from Ubiquity ARE NOT..

5442-s3 have very good image processing which process data from very good image sensor..

I hope firmware corrects all issues folks here including myself have pointed out. I would like to see what the raw data is like between the 5442-S3 and the new TIOC and PRO models
If it boils down to SoC, could this have anything to do with Dahua switch in SoC for cheaper SoC's over the past a couple of years?

 

[Arjun]

@Arjun That's because the IPC-Color4M-T cameras are trash, and EmpireTech is actively hiding bad reviews of them on their site (you really ought to leave another one, with multiple image comparisons like the above). Should've returned it and bought two of the vastly superior $95 IPC-Color4M-TL cameras instead! Just be aware that the IPC-Color4M-TL has a few quirks that need to be adjusted for to maximize performance:

• It uses a 1440p sensor despite having a 1520p encoding option, so be sure to run it at 1440p resolution to avoid blurriness and grid-like interpolation artifacts.
• For best image, start with Self-adaptive mode, Standard style, Saturation=55, Sharpness=25, Gamma=45, and 2D NR Level=15. Note that after setting these and exiting out of the camera UI, you will need to fully reboot the camera from Blue Iris due to the widespread firmware bug that reverts the 2D NR Level back to 50 upon closing the webpage (it'll reload your set value upon reboot). Note that if you fail to drop 2D NR Level, the camera will have poor image clarity; and if you fail to drop Sharpness, the camera will have an almost embossed image—the defaults really do it no justice!
• If you find the nighttime image too bright, I found it can be darkened by setting Illumination Overexposure Remover to Number Plate Priority. Note that this also causes the camera to flash sometimes, particularly around moving vehicles. While I don't care for the flashing, it does make it capture sharp plate images and choose a reasonable image brightness at night.

With these settings applied, I believe you will find the IPC-Color4M-TL will take pictures just as clear and vibrant as the 5442, with less noise and significantly better nighttime performance. It feels like the "little brother" to the IPC-Color4K-T and is now my "go to" camera unless visible illumination at night is undesirable. Too bad they don't come in a 6mm variety or varifocal!

I remember you writing about the TL model before, I still find it mind-boggling. I'm hesitant to spend more money until these new models reach a resolution and when Dahua admits their new models are not up to par as their previous models. Isn't the outer shell plastic of the TL though? I did your picture comparison on another thread and agree that the TL outperforms the more expensive models. $95 is a damn good price. Also, I wholeheartedly "hate" the new mounting system. Its a major downgrade.

 

[Techie007L]

@Arjun Sadly yes, the housing of the IPC-Color4M-TL camera is plastic, just like the cheap Amcrest IP5M-T1179 cameras. I had concerns that someone could reach up and just twist them in the absence of the locking screw, but I haven't had any problems with the ones I've installed so far (starting over a year ago). I figured I could try various mitigations if it became an issue (drive my own locking screw, add rubber, superglue the plastic, etc). You can hang from the old 5442's and they'd stay put if sufficiently tightened down and mounted to solid wood with 3-4" deck screws!

I wish EmpireTech would list the IPC-Color4M-TL on Amazon so I could positively review it on a reputable site that doesn't remove reviews (my review of the IPC-Color4M-T can be found there with a few comparison images; I would have done more testing but I was busy and realized I needed to return it right away if I was to get a refund). Perhaps I'll purchase the IPC-Color4M-TZ sometime and see how it does—I bet it's the same junk camera (uses the same firmware file) but EmpireTech assures me it's not. If not, great! But if so, that'll give me another month to explore the flaws of that chipset/sensor combo. It'll also give me the ability to test again with firmware that's a year newer to see if they've improved the image quality any.

 

[tigerwillow1]

The question isnt the FW - its the sensor

I've thought that the lousy IVS detection in the IL was because of firmware. I've got no complaints with the IL's visible image. Haven't seen one case where the S3 beats it, many cases where it beats the S3.