DXO PL6 strange behaviour

Thanks. I can flip between them in the DPR viewer and see that the main image is almost unaffected. The two houses in the lower corners change slightly. Vignetting?
I don't know — I didn't notice any differences when I flipped between them in Affinity.
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.

9cc6b12f6b9242e8a44713b416445850.jpg

2360744c0ade4ef8891a37549c1e24a1.jpg
 
Last edited:
Is there an official list of supported GPU? I've seen a list with "deep prime" in mind...
That's the only one I know of. See my screen shot of the Preferences > Performance page.
It would be very unusual to propose software that works oddly in particular configurations. Slow acting or "grayed-out" function is reasonable solution, but "odd results"? In a commercial product? Doesn't sound good...
The warning is that 'errors' are a possible consequence with partially supported hardware.
From their faq:

'For systems with older embedded graphics cards, it's possible that GPU acceleration could cause stability problems.

If so, please deactivate GPU acceleration and OpenCL (in case they are activated) in DxO PhotoLab's preferences."

No word about program's weird behaviour.
Did you not see my screen shot? Here it is again:

4a653ea6f54c41ae8d98dfee07dd5a4f.jpg
BTW, I think that this bug will reveal in each system, no matter if it is made of recommended elements or not.
We already know that two other experienced PhotoLab users here have said they can't duplicate the issue. This behavior (on my system) is seen in PL5, which was released two years ago. If something this obvious is widespread among all users, we should have seen other reports before now.
I checked this on PL5 too, but on the new system only. And did not noticed this bug earlier (playing with photography occasionally, if truth to be told) on any of PL versions. The bug became visible with particular set of images, that required quite a lot of adjustments involving high clearview settings.
This is probably something overlooked by the developers, and probably it requires deep code modifications to get rid of.
I would like you to just check the Preferences > Performance page and see if your GPU is marked with an asterisk. The goal is to find any unusual traits among systems that have the issue.
On my newer system bug "works" with and without GPU enabled. I did not checked it on the older system.

--
Marcin_3M
 
Just to clarify, here's a very simple way to see the issue. Open an image and apply global ClearView Plus at +100. Then apply a local adjustment somewhere using a brush. Change the Exposure setting in the local adjustment to +4.00. Make no other changes. Simply by showing and hiding that local adjustment, the entire image looks different.
This procedure is reproducible on my images. The whole image turns "bluer" when the local adjustment is visible.

Mask hidden
Mask hidden

Mask shown
Mask shown

I also submitted a support request to DxO.

--
Annie (She/her)
https://smallpotato.photography
 
Last edited:
One more example.

Here is test image:

378e8f4afe304eed86f8d86a027fc490.jpg

And here are applied modifications. Last image shows local settings:

6584dc5ab449402aa12aa75ec9dbbd80.jpg

d9c3b03827a94647b7a15729d8547333.jpg

dd0f7fbfe3574b62a8baf33a6a761c4e.jpg

Three above images are the images with local adjustments enabled. Local selection is done with a simple brush.

And here it is how it looks with local adjustments off:

1cf2e45807fb45a9a7a211ad33c2f569.jpg

Anyone can download source image (the first image from the series) and try to replicate bug. It "works" with jpg too.

--
Marcin_3M
 
Last edited:
Just to clarify, here's a very simple way to see the issue. Open an image and apply global ClearView Plus at +100. Then apply a local adjustment somewhere using a brush. Change the Exposure setting in the local adjustment to +4.00. Make no other changes. Simply by showing and hiding that local adjustment, the entire image looks different.
This procedure is reproducible on my images. The whole image turns "bluer" when the local adjustment is visible.

Mask hidden
Mask hidden

Mask shown
Mask shown

I also submitted a support request to DxO.
I can't wait to hear their responses. Hopefully you'll have better luck.

--
Marcin_3M
 
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.

LA off
LA off

Stronger LA on. The brightness of all other bright areas is reduced very slightly.
Stronger LA on. The brightness of all other bright areas is reduced very slightly.
 
I would like you to just check the Preferences > Performance page and see if your GPU is marked with an asterisk. The goal is to find any unusual traits among systems that have the issue.
On my newer system bug "works" with and without GPU enabled. I did not checked it on the older system.
You have repeatedly avoided complying with my simple request. Why?
 
Last edited:
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.
So it's fair to say that some anomaly does exist with your system, although it's much less apparent than it is with mine.

I know you normally use a custom preset. Maybe you can try with the No Correction preset applied just to limit any other variables.
 
Just to clarify, here's a very simple way to see the issue. Open an image and apply global ClearView Plus at +100. Then apply a local adjustment somewhere using a brush. Change the Exposure setting in the local adjustment to +4.00. Make no other changes. Simply by showing and hiding that local adjustment, the entire image looks different.
This procedure is reproducible on my images. The whole image turns "bluer" when the local adjustment is visible.
Thanks, that's a pretty significant change. Up until now, you and Digital Nigel were reporting no issues. I guess this is a better test.

Can you check by looking here to see if your GPU is fully or only partially supported?

4a653ea6f54c41ae8d98dfee07dd5a4f.jpg
 
Last edited:
I would like you to just check the Preferences > Performance page and see if your GPU is marked with an asterisk. The goal is to find any unusual traits among systems that have the issue.
On my newer system bug "works" with and without GPU enabled. I did not checked it on the older system.
You have repeatedly avoided complying with my simple request. Why?
Sorry, I did not get it. I'm not a native speaker :)

Here You are:

582b4e9a18fe47249001c445e50f9e29.jpg

005579d5c31d4886873da5cc000aca71.jpg

When I point at the exclamation mark it shows following:

cf920d6cb3814a799b03c2401d801bbb.jpg

Displayed statement is not correct.

I hope this is what You have asked for (I have PL6) ?

--
Marcin_3M
 
Last edited:
The effect depends in some way on the content of the image. This may be because the depth of the Clearview effect depends on certain image characteristics.

At the end of the thread I posted another test (link: https://www.dpreview.com/forums/post/67339176) with downloadable "synthetic" image. Only a few settings are used, so they are easy to reproduce. Maybe it is worth to check it in Your system?
 
I would like you to just check the Preferences > Performance page and see if your GPU is marked with an asterisk. The goal is to find any unusual traits among systems that have the issue.
On my newer system bug "works" with and without GPU enabled. I did not checked it on the older system.
You have repeatedly avoided complying with my simple request. Why?
Sorry, I did not get it. I'm not a native speaker :)

Here You are:

005579d5c31d4886873da5cc000aca71.jpg

When I point at the exclamation mark it shows following:

cf920d6cb3814a799b03c2401d801bbb.jpg

Displayed statement is not correct.

I hope this is what You have asked for (I have PL6) ?
No. This is what I'm asking for:

4a653ea6f54c41ae8d98dfee07dd5a4f.jpg

You will only see that warning in that particular place ... and you will only see the asterisk next to your GPU if it is not fully supported. I know this is in regard to DeepPRIME, but it still could indicate that 'errors' might result in other features.
 
Lol, I've edited my post while You have answered it, so You did not noticed this image (which now is in place):



0a953158d8034e0c8c1ab41857ad930a.jpg



--
Marcin_3M
 
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.
So it's fair to say that some anomaly does exist with your system, although it's much less apparent than it is with mine.

I know you normally use a custom preset. Maybe you can try with the No Correction preset applied just to limit any other variables.
I've tried it, and the effect is the same.

Reducing the opacity of the local adjustment reduces the effect, which is probably why most people don't see it — I would never use such an extreme local adjustment. At the normal levels I use, the effect is too small to be seen, or to matter.
 
Lol, I've edited my post while You have answered it, so You did not noticed this image (which now is in place):

0a953158d8034e0c8c1ab41857ad930a.jpg
Thank you. Then it appears that the effect can be fairly easily seen on some systems even if the GPU is fully supported.
 
Lol, I've edited my post while You have answered it, so You did not noticed this image (which now is in place):

0a953158d8034e0c8c1ab41857ad930a.jpg
Thank you. Then it appears that the effect can be fairly easily seen on some systems even if the GPU is fully supported.
So, we wait for DXO to step in...

--
Marcin_3M
 
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.
So it's fair to say that some anomaly does exist with your system, although it's much less apparent than it is with mine.

I know you normally use a custom preset. Maybe you can try with the No Correction preset applied just to limit any other variables.
After some more experiments, I think I've pinned it down to how ClearView works.

It seems to do some sort of averaging across the image to produce a balanced effect. So, if you have a large local adjustment that does dramatic things in part of the image, CV adjusts the rest of the image to balance it (so a very bright LA area causes a slight darkening in the rest of the image). This applies even if the CVP is applied via another local adjustment rather than globally.

If CVP is used at typical levels (ie, much less than 100), the effect is too mild to be noticeable.

What isn't clear is whether this is a carefully designed effect that's been deliberately programmed into the ClearView algorithm, or a bug.
 
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.
So it's fair to say that some anomaly does exist with your system, although it's much less apparent than it is with mine.

I know you normally use a custom preset. Maybe you can try with the No Correction preset applied just to limit any other variables.
After some more experiments, I think I've pinned it down to how ClearView works.

It seems to do some sort of averaging across the image to produce a balanced effect. So, if you have a large local adjustment that does dramatic things in part of the image, CV adjusts the rest of the image to balance it (so a very bright LA area causes a slight darkening in the rest of the image). This applies even if the CVP is applied via another local adjustment rather than globally.

If CVP is used at typical levels (ie, much less than 100), the effect is too mild to be noticeable.

What isn't clear is whether this is a carefully designed effect that's been deliberately programmed into the ClearView algorithm, or a bug.
The other part that isn't clear is why some systems show a very distinct difference while in others the difference is barely perceptible. Or is that a consequence of the content of the image? So far, I don't remember any of us running the same tests using the same source image on our own systems.
 
Last edited:
Isolate those corners and check again.

Actually, it's not just vignetting. When the house and the roof of the shed get brighter, the brown foliage gets darker.
Yes, agreed. I've done another test, with a stronger LA effect, and I can see that the effect, though slight, does affect brighter areas across the image.
So it's fair to say that some anomaly does exist with your system, although it's much less apparent than it is with mine.

I know you normally use a custom preset. Maybe you can try with the No Correction preset applied just to limit any other variables.
After some more experiments, I think I've pinned it down to how ClearView works.

It seems to do some sort of averaging across the image to produce a balanced effect. So, if you have a large local adjustment that does dramatic things in part of the image, CV adjusts the rest of the image to balance it (so a very bright LA area causes a slight darkening in the rest of the image). This applies even if the CVP is applied via another local adjustment rather than globally.

If CVP is used at typical levels (ie, much less than 100), the effect is too mild to be noticeable.

What isn't clear is whether this is a carefully designed effect that's been deliberately programmed into the ClearView algorithm, or a bug.
The other part that isn't clear is why some systems show a very distinct difference while in others the difference is barely perceptible. Or is that a consequence of the content of the image? So far, I don't remember any of us running the same tests using the same source image on our own systems.
Yes, that's why I asked the OP to share a raw+dop file to try on other machines, but he refused. I'm now coming to suspect that computer hardware or PL version or camera model isn't a factor: it's just how CVP works, probably by design, but perhaps accidentally.
 
... What isn't clear is whether this is a carefully designed effect that's been deliberately programmed into the ClearView algorithm, or a bug.
The other part that isn't clear is why some systems show a very distinct difference while in others the difference is barely perceptible. Or is that a consequence of the content of the image? So far, I don't remember any of us running the same tests using the same source image on our own systems.
Yes, that's why I asked the OP to share a raw+dop file to try on other machines, but he refused. I'm now coming to suspect that computer hardware or PL version or camera model isn't a factor: it's just how CVP works, probably by design, but perhaps accidentally.
It doesn't have to be RAW, does it? Let's try a DPR source image.

DPR source image
DPR source image

Apply these settings:
  • DxO Preset: No correction
  • Color Space: Wide Gamut with Generic Rendering (should be the default with v6/ v7)
  • Global CVP: +100
  • Local adjustment brush with Exposure set to +4.00
Here's what I get:

Brush selection enabled
Brush selection enabled

Brush selection disabled
Brush selection disabled
 
Last edited:

Keyboard shortcuts

Back
Top