Re: [darktable-user] Base curve and white balance issue with DT 3.0

[jys]

On Tue, Jan 14, 2020, at 06:31, Šarūnas wrote:

> I hear what you are saying. I use Olympus cameras since one of their
> first digital SLR, E-500, and like Olympus's out of camera (OOC) color
> rendition very much. It would be my preferred starting point while
> working on RAW.

I'm coming from a similar place, with some reservations... I've always been 
slightly annoyed with the tendency to blow out reds into magenta and turn sky a 
little too cyan, but in general I think Olympus rendering has been pretty good, 
at least for images which benefit from the "punchy" contrast (some obviously 
don't). With the original filmic module, I was able to closely match the curve 
to the camera rending with the help of an IT-8 chart and a few beers. The 
settings were fairly extreme, which somewhat defeated the purpose of having a 
default with nice parametric adjustment available... but it was possible.

The new filmic RGB features a custom spline which enforces more linearity in 
the midrange, and while I understand the idea behind it, I think it may be the 
source of some people's frustrations. Simply put, the "look" it provides is not 
one that everyone likes for everything, even with increases to local/global 
contrast, etc. My understanding is that one of the problems with the old 
version was a tendency towards oscillations, although I didn't experience this 
myself. If this could be mitigated to a reasonable extent, then adding an 
option to the new filmic to use something more like the old spline might be a 
real "crowd pleaser", and increase the general acceptance of the module... and 
the underlying idea of the module is a really good one!

So, because I really like the idea of the parametric controls which filmic 
offers, my own approach lately has been a hybrid one: using a target shot and a 
variety of test images, I tried to find a generally sane default preset for 
filmic RGB without any extreme settings. It gives "reasonable" renderings of my 
camera's RAWs, even if they don't resemble the camera's rendering. Then, using 
the same corpus of test images, I manually created an "extra" curve to add a 
"knee" to filmic's relative linearity, bringing things close enough (not 
*quite* as "punchy") to camera rendering for my tastes. Applying both of these 
by default, I can simply disable the extra curve for the images which are 
better suited to "plain filmic"... and either way have some parametric control 
for adjustment. It seems to work well so far, but I'm absolutely not implying 
that this is a "correct" approach. :-)

The attached style is what I'm using as of this moment; it may work well with 
Olympus 16MP sensor cameras generally, or may not, but it's an example of what 
I'm describing, for whatever that's worth. I also so some color profiling with 
the color LUT module, but it's probably very camera-specific, hence not 
included in the style. So, that's my two cents!

-- 
jys

darktable user mailing list
to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org



filmic+curve_olympus-like.dtstyle
Description: Binary data

Re: [darktable-user] Base curve and white balance issue with DT 3.0

[Terry Pinfold]

Hope my email gets through. I have invested time learning RGB Filmic and it is 
very good at cature the maximum dynamic range. I find when I finish with RGB 
Filmic I need to use the tone curve to just push up the contrast a little. 
However, unless I can create a reliable preset (not sure if I could) I would 
not want to have to use Filmic on every image. I like the way RawTherapee 
references the embedded JPEG to replicate what the camera would produce. I even 
tested Lightroom, Darktable and Rawtherapee with a Nikon Raw file where 
D-lighting has been applied in camera to brighten the shadows. RawTherapee was 
the only one that factored in the D-Lighting change.

I would like to use some preset base curves for my average shots that do not 
require Filmic. I will probably invest some time and create some styles for 
various genre of images such as Sunny Day Landscape, Overcast Sky Landscape, 
Portraits etc. That being said Filmic is a great tool but not for everyday 
images.

Dr Terry Pinfold
Cytometry & Histology Lab Manager
University of Tasmania
17 Liverpool St, Hobart, 7000
Ph 6226 4846 or 0408 699053

From: Viktors Krasovskis 
Sent: Tuesday, 14 January 2020 11:35 PM
To: Remco Viëtor 
Cc: Darktable-users list 
Subject: Re: [darktable-user] Base curve and white balance issue with DT 3.0

Hi Remco. I understand the Darktable developers point to push the ultimate raw 
interpretation and introduce raw processing workflow that squeezes out most 
data into richer result. It's cool and makes sense in theory or if you don't 
shoot a lot and have time to develop the raw files. I don't know how Sony 
renders it's JPEGs, though Nikon, Fuji and Canon do the job quite well in terms 
to provide universal preview for many scene types. When photographer takes 
pictures he checks the result in EVF or on camera's display and makes the 
exposure, light, subject and scene adjustments by the JPEG preview. After some 
practice the photographer learns to know his camera and it's possible to 
predict by JPEG preview how far we can go with under/overexposing the image. If 
I do some shooting on location where I often change the lighting proportions in 
my photos then it's a pain to adjust filmic RGB module sliders for each 
pictures because they all are so different. Well and the filmic presets do not 
help much. Ok, I try to process each image with filmic not looking to camera 
JPEGs to be unbiased in my editing. And later, when i compare the results with 
camera JPEGs I see that JPEGs look more natural and more richer in shadows and 
middle tones, the skin tones are just right in most pictures by default and I 
could fix the others just by adjusting the white balance. Maybe sometimes the 
camera JPEGs are less saturated, underexposed or have shifted white balance. 
But that's easy to fix in DT 2.6.x since the RAW rendering starting point is 
close to the camera JPEG and represents the preview which I created on the 
field with my adjustments. Now the DT 3.0 brakes this solid workflow, I can't 
rely on my camera's preview while shooting because now I know that the starting 
point on darktable will look different. That's why I downgraded back to 2.6.x. 
DT 2.6.x is a great tool and I'm afraid that the development of this version 
will phase out and it won't get new camera support, new modules and bugfixes.
Darktable 3.0 would be a fantastic tool if it would allow the users to choose 
between the classic and advanced workflows. That would speed up editing where 
camera manufacturer raw interpretation worked well and for more complicated 
cases filmic RGB module would be a nice help for experienced users. Most users 
in most cases need that their raw files are rendered well enough and the 
rendering should be predictive. Only geeks and some others (the minority) needs 
the ultimate raw rendering which takes more time and in practice is less 
universal. Filmic RGB is great for landscapes and architecture where the 
camera's dynamic range is on it's limits, however Nikon base curve presents 
give far better results on human portraits on DT 2.6.x by default. What do 
others think and do they have similar experience?

вт, 14 янв. 2020 г. в 12:49, Remco Viëtor 
mailto:remco.vie...@wanadoo.fr>>:
On lundi 13 janvier 2020 15:28:04 CET Timur Irikovich Davletshin wrote:
> Hi Viktors,
>
> There are two problems in your case:
>
> 1. Nikon-like alternative basecurve turned just Nikon-like.
> 2. New color preservation settings which messed highlights and colors.
>
> First issue: I had similar problem in the past. I believe there was
> problem with Exif interpretation. E.g. image information used to show
> Nikon d7100 but now it shows NIKON D7100. Can you try to reimport some
> file and check it in image information?
>
> Second issue: I addressed it in the past
> https://github.com/darktable-org/darktable/issues/3693
>  and
> https://github.com/darktable-

Re: [darktable-user] Base curve and white balance issue with DT 3.0

[Marcus Amorim]

Hi Viktors!

I've also been having a hard time using the new filmic rgb module. I can't
get neither a better image than my Fuji X70's JPEG, as I could on DT 2.6.2.
I found the original filmic module easier, more straightforward to use.

Yesterday I tried the LUTs that Guido indicated. Even being for x-trans
III, they worked perfectly with my x-trans II RAW files. I think I'll leave
filmic rgb and start my workflow with these profiles for Fuji. I'm just not
sure if 3d-luts is the ideal option for my workflow. Any thoughts?

Marcus

Em ter., 14 de jan. de 2020 às 09:36, Viktors Krasovskis <
viktors.krasovs...@gmail.com> escreveu:

> Hi Remco. I understand the Darktable developers point to push the ultimate
> raw interpretation and introduce raw processing workflow that squeezes out
> most data into richer result. It's cool and makes sense in theory or if you
> don't shoot a lot and have time to develop the raw files. I don't know how
> Sony renders it's JPEGs, though Nikon, Fuji and Canon do the job quite
> well in terms to provide universal preview for many scene types. When
> photographer takes pictures he checks the result in EVF or on camera's
> display and makes the exposure, light, subject and scene adjustments by the
> JPEG preview. After some practice the photographer learns to know his
> camera and it's possible to predict by JPEG preview how far we can go with
> under/overexposing the image. If I do some shooting on location where I
> often change the lighting proportions in my photos then it's a pain to
> adjust filmic RGB module sliders for each pictures because they all are so
> different. Well and the filmic presets do not help much. Ok, I try to
> process each image with filmic not looking to camera JPEGs to be unbiased
> in my editing. And later, when i compare the results with camera JPEGs I
> see that JPEGs look more natural and more richer in shadows and middle
> tones, the skin tones are just right in most pictures by default and I
> could fix the others just by adjusting the white balance. Maybe sometimes
> the camera JPEGs are less saturated, underexposed or have shifted white
> balance. But that's easy to fix in DT 2.6.x since the RAW rendering
> starting point is close to the camera JPEG and represents the preview which
> I created on the field with my adjustments. Now the DT 3.0 brakes this
> solid workflow, I can't rely on my camera's preview while shooting because
> now I know that the starting point on darktable will look different. That's
> why I downgraded back to 2.6.x. DT 2.6.x is a great tool and I'm afraid
> that the development of this version will phase out and it won't get new
> camera support, new modules and bugfixes.
> Darktable 3.0 would be a fantastic tool if it would allow the users to
> choose between the classic and advanced workflows. That would speed up
> editing where camera manufacturer raw interpretation worked well and for
> more complicated cases filmic RGB module would be a nice help for
> experienced users. Most users in most cases need that their raw files are
> rendered well enough and the rendering should be predictive. Only geeks and
> some others (the minority) needs the ultimate raw rendering which takes
> more time and in practice is less universal. Filmic RGB is great for
> landscapes and architecture where the camera's dynamic range is on it's
> limits, however Nikon base curve presents give far better results on human
> portraits on DT 2.6.x by default. What do others think and do they have
> similar experience?
>
> вт, 14 янв. 2020 г. в 12:49, Remco Viëtor :
>
>> On lundi 13 janvier 2020 15:28:04 CET Timur Irikovich Davletshin wrote:
>> > Hi Viktors,
>> >
>> > There are two problems in your case:
>> >
>> > 1. Nikon-like alternative basecurve turned just Nikon-like.
>> > 2. New color preservation settings which messed highlights and colors.
>> >
>> > First issue: I had similar problem in the past. I believe there was
>> > problem with Exif interpretation. E.g. image information used to show
>> > Nikon d7100 but now it shows NIKON D7100. Can you try to reimport some
>> > file and check it in image information?
>> >
>> > Second issue: I addressed it in the past
>> > https://github.com/darktable-org/darktable/issues/3693 and
>> > https://github.com/darktable-org/darktable/issues/3677 — so nothing to
>> > do.
>> >
>> > Timur.
>> >
>> > On Sun, 2020-01-12 at 13:04 +0200, Viktors Krasovskis wrote:
>> > > Hi. The DT 3.0 is nice. However I noticed a serious problem when
>> > > working with my Nikon D7200 RAW files. The DT 2.6.2 version I was
>> > > using did automatically apply the D7200 base curve preset and the
>> > > initial rendering of the RAW file looked quite similar to camera's
>> > > JPEG file (tones, saturation, white balance and the exposure). It was
>> > > a good starting point for my editing. Now the DT 3.0 applies a wrong
>> > > base curve preset (nikon like) and when I choose the D7200 base curve
>> > > preset then the images looks desatu

Re: [darktable-user] ppa

[Michael]

 is Pascal okay? I've been waiting for the new darktable from him for a
 while now and am getting concerned for him.
 --
 :-)~MIKE~(-:



-- 
:-)~MIKE~(-:

darktable user mailing list
to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org

Re: [darktable-user] 3.0 (for Guido)

[Hervé Sainct]

Thank you Guido, that's perfect.

In addition, I'm now more motivated to switch to Debian 10, which is
still adventurous for me ;-)

Thanks again,

H.

Le 13/01/2020 à 21:39, Guido Scholz a écrit :
> Am Fri, 10. Jan 2020 um 18:36:57 +0100 schrieb Hervé Sainct:
>
> Hi Hervé,
>
>> Guido, I too was extremely interested, and (being quite the newbie) I
>> tried.
>>
>> Everything goes well until the 'fakeroot' build command, where I get
>>
>> (a) a long series of warning about Lua* and then
>>
>> (b) an error 2 fatal failure as 'the recipe for target "
>> override_dh_auto_configure " has failed'
>>
>> I am unable to interpret what this override failure means... If you have
>> any advice...
> sorry for the delay but I had to try some things first. So my decription
> above worked on a Ubuntu 19.10 but obviously not on a 18.04 or Mint
> 19.3. Attached you find an other control file set, which does the job on
> those distributions and should work on your 9.11 as well.
>
> To get things done safe this file (darktable_3.0.0-2.debian.tar.xz) to
> your ~/src/debian directory and follow this description:
>
>
> # goto directory
> cd ~/src/debian
>
> # remove original control file set
> rm darktable_3.0.0-1.debian.tar.xz
>
> # remove used source code directory
> rm -rf darktable-3.0.0
>
> # freshly restart the building process
> tar -xf darktable_3.0.0.orig.tar.xz
> cd darktable-3.0.0
> tar -xf ../darktable_3.0.0-2.debian.tar.xz
>
> # build package
> dpkg-buildpackage -rfakeroot  -uc -us
> cd ..
>
> #install package
> sudo dpkg -i darktable_3.0.0-1_amd64.deb
>
>
> Guido
>



signature.asc
Description: OpenPGP digital signature

Re: [darktable-user] Base curve and white balance issue with DT 3.0

[Šarūnas]

On 1/14/20 7:35 AM, Viktors Krasovskis wrote:
> ...
> Darktable 3.0 would be a fantastic tool if it would allow the users to
> choose between the classic and advanced workflows. That would speed up
> editing where camera manufacturer raw interpretation worked well and for
> more complicated cases filmic RGB module would be a nice help for
> experienced users. Most users in most cases need that their raw files
> are rendered well enough and the rendering should be predictive. Only
> geeks and some others (the minority) needs the ultimate raw rendering
> which takes more time and in practice is less universal. Filmic RGB is
> great for landscapes and architecture where the camera's dynamic range
> is on it's limits, however Nikon base curve presents give far better
> results on human portraits on DT 2.6.x by default. What do others think
> and do they have similar experience?

I hear what you are saying. I use Olympus cameras since one of their
first digital SLR, E-500, and like Olympus's out of camera (OOC) color
rendition very much. It would be my preferred starting point while
working on RAW. However, none of the RAW conversion software (except
Olympus's own) comes close to OOC colors in some universal/preset way.
Base curves for Olympus didn't do that either, and I stopped using them
in darktable 2.x a while ago. The closest I was able to get to OOC in
some automated way was by creating a preset with a camera profile,
produced by photographing IT8 color target. Here is the latest write-up
on the technique by Andreas Schneider:

https://pixls.us/articles/profiling-a-camera-with-darktable-chart

The resulting preset worked quite well for some exposures/lighting
conditions, but for quite a few it didn't, so in the end I gave up on
that either.

-- 
Šarūnas Burdulis
math.dartmouth.edu/~sarunas

· https://useplaintext.email ·




signature.asc
Description: OpenPGP digital signature

Re: [darktable-user] Base curve and white balance issue with DT 3.0

[Viktors Krasovskis]

Hi Remco. I understand the Darktable developers point to push the ultimate
raw interpretation and introduce raw processing workflow that squeezes out
most data into richer result. It's cool and makes sense in theory or if you
don't shoot a lot and have time to develop the raw files. I don't know how
Sony renders it's JPEGs, though Nikon, Fuji and Canon do the job quite well
in terms to provide universal preview for many scene types. When
photographer takes pictures he checks the result in EVF or on camera's
display and makes the exposure, light, subject and scene adjustments by the
JPEG preview. After some practice the photographer learns to know his
camera and it's possible to predict by JPEG preview how far we can go with
under/overexposing the image. If I do some shooting on location where I
often change the lighting proportions in my photos then it's a pain to
adjust filmic RGB module sliders for each pictures because they all are so
different. Well and the filmic presets do not help much. Ok, I try to
process each image with filmic not looking to camera JPEGs to be unbiased
in my editing. And later, when i compare the results with camera JPEGs I
see that JPEGs look more natural and more richer in shadows and middle
tones, the skin tones are just right in most pictures by default and I
could fix the others just by adjusting the white balance. Maybe sometimes
the camera JPEGs are less saturated, underexposed or have shifted white
balance. But that's easy to fix in DT 2.6.x since the RAW rendering
starting point is close to the camera JPEG and represents the preview which
I created on the field with my adjustments. Now the DT 3.0 brakes this
solid workflow, I can't rely on my camera's preview while shooting because
now I know that the starting point on darktable will look different. That's
why I downgraded back to 2.6.x. DT 2.6.x is a great tool and I'm afraid
that the development of this version will phase out and it won't get new
camera support, new modules and bugfixes.
Darktable 3.0 would be a fantastic tool if it would allow the users to
choose between the classic and advanced workflows. That would speed up
editing where camera manufacturer raw interpretation worked well and for
more complicated cases filmic RGB module would be a nice help for
experienced users. Most users in most cases need that their raw files are
rendered well enough and the rendering should be predictive. Only geeks and
some others (the minority) needs the ultimate raw rendering which takes
more time and in practice is less universal. Filmic RGB is great for
landscapes and architecture where the camera's dynamic range is on it's
limits, however Nikon base curve presents give far better results on human
portraits on DT 2.6.x by default. What do others think and do they have
similar experience?

вт, 14 янв. 2020 г. в 12:49, Remco Viëtor :

> On lundi 13 janvier 2020 15:28:04 CET Timur Irikovich Davletshin wrote:
> > Hi Viktors,
> >
> > There are two problems in your case:
> >
> > 1. Nikon-like alternative basecurve turned just Nikon-like.
> > 2. New color preservation settings which messed highlights and colors.
> >
> > First issue: I had similar problem in the past. I believe there was
> > problem with Exif interpretation. E.g. image information used to show
> > Nikon d7100 but now it shows NIKON D7100. Can you try to reimport some
> > file and check it in image information?
> >
> > Second issue: I addressed it in the past
> > https://github.com/darktable-org/darktable/issues/3693 and
> > https://github.com/darktable-org/darktable/issues/3677 — so nothing to
> > do.
> >
> > Timur.
> >
> > On Sun, 2020-01-12 at 13:04 +0200, Viktors Krasovskis wrote:
> > > Hi. The DT 3.0 is nice. However I noticed a serious problem when
> > > working with my Nikon D7200 RAW files. The DT 2.6.2 version I was
> > > using did automatically apply the D7200 base curve preset and the
> > > initial rendering of the RAW file looked quite similar to camera's
> > > JPEG file (tones, saturation, white balance and the exposure). It was
> > > a good starting point for my editing. Now the DT 3.0 applies a wrong
> > > base curve preset (nikon like) and when I choose the D7200 base curve
> > > preset then the images looks desaturated, underexposed and less
> > > detailed. I tried to fix this with the exposure, white balance,
> > > saturation and contrast sliders and I still can't get a similar look
> > > like in camera's JPEG, the skin tones look weird, the shadows are too
> > > dark, bet when I raise them I loose contrast. In other words the
> > > colors are not so natural like the were rendered in DT 2.6.2. I've
> > > also tried the filmic RGB module (with the base curve and without),
> > > still can't adjust the image better as it was done by default in DT
> > > 2.6.2. What I'm doing wrong? Is it a bug? I can provide my Nikon RAW
> > > and JPEG files to compare.
> > >
>
> One thing to keep in mind: the basecurves give *one* interpretation of the
> raw
> data, n

Re: [darktable-user] Base curve and white balance issue with DT 3.0

[Remco Viëtor]

On lundi 13 janvier 2020 15:28:04 CET Timur Irikovich Davletshin wrote:
> Hi Viktors,
> 
> There are two problems in your case:
> 
> 1. Nikon-like alternative basecurve turned just Nikon-like.
> 2. New color preservation settings which messed highlights and colors.
> 
> First issue: I had similar problem in the past. I believe there was
> problem with Exif interpretation. E.g. image information used to show
> Nikon d7100 but now it shows NIKON D7100. Can you try to reimport some
> file and check it in image information?
> 
> Second issue: I addressed it in the past
> https://github.com/darktable-org/darktable/issues/3693 and
> https://github.com/darktable-org/darktable/issues/3677 — so nothing to
> do.
> 
> Timur.
> 
> On Sun, 2020-01-12 at 13:04 +0200, Viktors Krasovskis wrote:
> > Hi. The DT 3.0 is nice. However I noticed a serious problem when
> > working with my Nikon D7200 RAW files. The DT 2.6.2 version I was
> > using did automatically apply the D7200 base curve preset and the
> > initial rendering of the RAW file looked quite similar to camera's
> > JPEG file (tones, saturation, white balance and the exposure). It was
> > a good starting point for my editing. Now the DT 3.0 applies a wrong
> > base curve preset (nikon like) and when I choose the D7200 base curve
> > preset then the images looks desaturated, underexposed and less
> > detailed. I tried to fix this with the exposure, white balance,
> > saturation and contrast sliders and I still can't get a similar look
> > like in camera's JPEG, the skin tones look weird, the shadows are too
> > dark, bet when I raise them I loose contrast. In other words the
> > colors are not so natural like the were rendered in DT 2.6.2. I've
> > also tried the filmic RGB module (with the base curve and without),
> > still can't adjust the image better as it was done by default in DT
> > 2.6.2. What I'm doing wrong? Is it a bug? I can provide my Nikon RAW
> > and JPEG files to compare.
> > 

One thing to keep in mind: the basecurves give *one* interpretation of the raw 
data, not *the* interpretation. 

And while it provides an easy starting point, there are disadvantages: I 
noticed that in my case, the automatically selected curve (Sony-like) threw 
away about 1 stop in the highlights. And there are more basecurves that behave 
that way.

So following the camera maker's taste has its issues...

Remco




darktable user mailing list
to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org