Agx terminology (UI)

Hello,
Personally, I am one of those who does not use the relative exposure sliders because I find it more difficult to adjust the contrast and its curve afterwards.
On the other hand, I find that the new button works much better than the auto-tune levels, so I would definitely keep it and maybe even add a keyboard shortcut.

I tried redoing several processes by resetting the AGX module, and with the new button, I tend to get a “softer” result. Here’s an example

Original edit


With the new button

Left = original

What is the difference between the old “Power” slider and the new “Brightness” slider? I really like that it brightens when you slide it to the right, but it seems to me that it reduces the contrast a little compared to the old “Power” slider. Is this possible, or is it an illusion?
Edit: It was indeed an illusion, see kofa’s reply below.

I would like to take this opportunity to thank you and all the Darktable developers for the quality of your work and your team spirit. Well done! Darktable 5.4 will benefit from a major upgrade.

Greetings,
Christian

6 Likes

Nothing. Only the value is inverted on the UI.

2 Likes

Thanks. You can also try the exposure auto-picker.
The camera icon just reads the in-camera exposure compensation, and the exposure module, but does not analyse the image. The pickers are based on the actual content and try to ‘fill the histogram’.

2 Likes

Thank you for your reply, :+1:

I find the pivot picker can be the most impactful. if you select an area that you want to have the most contrast then you have a lot less tweaking to do and I find it a much better starting point than auto on the whole image or manually trying to hit it without the picker…

2 Likes

I was interested to see the way Boris used this on a face in his video. I’m often now doing a similar thing. I’m confused about the differences between input and output (of course I can see how the point moves on the curve graph) but sill managing to tweak to satisfaction.

Very much looking forward to seeing these changes in master branch :slight_smile:

The input is the region around which you want to set the contrast, or whose brightness you want to adjust. The output is the brightness you assign to that input in the processed image.

2 Likes

There was a great example recently. There was a playraw with two young men at a wedding.taking a silly picture in a box. I think the query was about wb. Nevertheless, if you enabled agx with defaults the young man on the right his face went to mush but just selecting his face and a touch of contrast and the image was perfect and just needed a little overall exposure bump…

1 Like

Brilliant, thank you.

I think my understanding was stumbling in that direction, but would might never have become so succinct.

Can I suggest that you add that to your documentation? I think it might help a lot of people.

1 Like

I tested this version on a series of low and high dynamic range images (directly into the sun). I almost never use the ‘white/black relative exposure’ pipettes; too often the s-shape of the curve is lost. So I find this ‘camera’ button useful, it leaves enough DR room to keep an s-shape but I fiddled around less with the shoulder and toe power/start buttons to get good results.
(I’m wondering with how many good ideas you’ll come up until Xmas :grinning:)

I also looked at the primaries section. For me the TRUE option is easier to use because the rotation in the vectorscope is the same as the slider color (cyan to cyan). With the FALSE option the vectorscope turns in the other direction (slider to cyan, colors in vectorscope to magenta).

I was even wondering whether the ‘after tone mapping’ part should keep the same slider colors as the ‘before tone mapping’ (like in the FALSE option) but with the ‘TRUE’ behavior, i.e. if you turned blue ‘after tm’ to cyan, blue also turns to cyan in the vectorscope. And if you want to ‘restore’ a color after a rotation in ‘before tm’ (e.g. you rotated blue to cyan in ‘before tm’) then you have to turn it back in the opposite direction (i.e. to magenta) in the ‘after tm’ to get the ‘original’ color back. I think that’s what Boris suggest above. But I could also easily live with the current TRUE option.

The camera button also only adjusts the relative black/white exposure. And keep the pivot on the diagonal will maintain the S-shape, as long as there is enough contrast, if that is desired (it’s not always strictly necessary).

Yes I’ve seen that. But it seems to make just make a subtle adjustment so that the handling of the contrast sliders become easier. At least that’s the impression I had from my limited testing (about 10 images).

1 Like

Sure. Thanks for sharing; that’s why I asked. :slight_smile: If it’s useful, it can stay.

No AgX

AgX default

camera

auto tune

Default: Shadows too soft
Camera: Perfect
Auto tune: Shadows too dark

all results are very good.

1 Like

No AgX


AgX default

Camera not working

Auto Tune

1 Like

I like all the changes you did. I need to further test the rotation sliders. My initial inclination is to use the “true option”.
I am not clear on the intend of the new exposure button. Is it to read the exposure for the exposure slider, camera setting, or both?
I guess the key word is the “initial exposure”. Since, the button reacted to each subsequent change in exposure slider module and is not dynamically react to the exposure slider and must clicked after each changes in exposure. In short, I am a bit puzzled about what it does or suppose to do. I looked at others use and test of it and seems it gets a better result. Is it a shortcut to one that can be done manually? I am all for it if it is a shortcut to a manual setting.

Mid grey has 18% reflectance. Cameras base their exposure measurements around that value.

However, cameras “underexpose”. If they didn’t do that, then the brightest they could represent would be 100% diffuse white (e.g. snow, or a very pure white piece of paper or piece of clothing) – all light sources, including bright sky, metallic reflections, sunlight or lamps, even coming through translucent glass, etc. would be totally blown out. This underexposed image is normally brightened using the in-camera tone curve to produce the JPG.

In darktable, you will see that the exposure module adds 0.7 EV by default. This is to fix that underexposure – darktable’s tone mappers (whether it’s filmic, sigmoid or now agx) keep mid tones (specifically, the 18% point) unchanged.

There is also a compensate camera exposure setting. If you intentionally dial in an exposure compensation of, say, 1 EV, the camera will underexpose by the set amount. The exposure module can automatically read that, and increase total exposure by the same amount. And, more recently, it can even read the underexposure values cameras use for ‘HDR’ modes.

The idea is that, after the exposure module (with the user input and calculated boost) midtones are where the user wants them to be. By finding out how much exposure boost was added using the exposure module, we can estimate what the brightest value in the image is. filmic uses that, so it does not always start with the +4 EV relative white exposure (4 EV above mid grey). sigmoid does not need it, as its curve never saturates. agx is like filmic in that regard, as it has hard cut-off limits.

This is an experiment, if it would make sense to read the exposure data, and behave like filmic does. Since I was not sure it would work, I left the initial exposure limits unchanged, and wired it to a button. Also, if you change exposure afterward agx is enabled, you can use the button to re-read the value.

4 Likes

Thank you for the explanation. A lot to digest. Let me see if I understood it correctly. There is default hard cut-off limit ( +4 EV ). The limits can be changed to become relative to exposure slider after using the button. If so why the camera icon on the button? or did I get it backward?

I was ignorent about that, I only knew sensor raw data stored in raw format for RGB channels. I did not know what does happen before or after that.

In that case it is one more useful additional tool.

It impressive and admirable that you put so much work into this module. I am getting persistently good results without the need to use many other modules.

I want to thank you for your patient answering mine as well as other users questions. DT users will be pleasantly will be surprised this December with the new release of DT with Agx.

3 Likes

There is a hard default cut-off at mid-grey +4EV in filmic (that is the default value of the white relative exposure slider), but filmic also reads the camera exposure compensation from the EXIF data, then adds 0.7 EV (the default value of the exposure boost in the exposure module), then multiplies that by 0.8, and adds the result to the original 4 EV. So, if there is no camera exposure compensation, you get 4.56 EV in filmic for the white relative exposure (4 + 0.8 * 0.7). If there is a 2 EV in-camera adjustment, you get 4 + 0.8 * (0.7 + 2) = 6.16 EV.
There is no such thing in agx, it starts with relative white exposure = 6.5 EV as the cut-off. You can:

  • set a different value manually (like in filmic)
  • use the picker to check image content and make sure the brightest value reaches 1 (like in filmic, but a bit different, as filmic has the notion of a ‘norm’, while agx is per-channel)
  • use the ‘camera’ icon to perform a calculation that’s similar to what filmic does.
2 Likes

Understood. Thx.
If you decided to keep ( I think you should ) then icon on the button maybe a bit misleading.