My impression is that whenever the conversation turns to code and math, many forum participants feel that the discussion is inappropriate, confusing or off-topic. I think this is quite understandable, since the forum’s main focus is on the art of photograpy, and the use of OSS to achieve that art.
Having said that, many of us are also interested in the underlying math, the algorithms, and the implementations that form the basis of OS processors. So it seems natural to me to carve out a space for those discussions, out of sight of forum members who are not interested.
Is adding a category a big burden on the moderators? I’d be glad to moderate that category if that would reduce the workload.
The Welcome page of the forum states at its mission statement " To provide tutorials, workflows and a showcase for high-quality photography and cinematography using Free/Open Source Software."
This discussion, (where much of it lies above my own non-technical/non-math incompetency level, but which I anyhow as a user find interesting as there is a lot to be learned from it), I most importantly think forms a basis for future tutorials as it obviously clarify matters of importance for processing methods that many have been wondering about.
Scrolling the thread it’s not immediately apparent what is “algorithms/math/code” …
The burden would be in splitting threads and/or moving them once a discussion gets “too technical”. Most start innocent enough, but end up needing a technical explaination for the why.
Here, I tried to avoid the technical details. But that wasn’t enough explanation.
And now we should move all technical discussions to a specialised place… Most discussions don’t start being technical, they become that way.
I’d say either Software or Processing should suffice as a category, no?
You can tag the topic as “algorithm” or “math” as well to provide additional context for users.
I think if it’s sufficiently tagged, then caveat emptor if they enter the topic and find it too technical.
Please keep in mind that according to the latest survey, about 40% of Darktable users have a master’s or PhD degree (which, of course, may not be in a math-intensive field).
You may not find these discussions interesting, but many participants do. I can’t speak for everyone else, but I have gained a lot from understanding the math behind some DT modules, that resulted in improved images. We all understand things differently.
Also, for this particular discussion, note that the tone mapping part of both sigmoid and filmic can be understood only high school math.
FWIW, I have no degree whatsoever and very little tertiary education, (and I know it shows!) yet I have nothing against the technical discussions. I can usually follow the discussion at least partly, and sometimes I’m intrigued enough to chase down a few rabbit holes until I understand it better.
It’s all fair play and one of the good things about this forum IMO.
What are you trying to visualise exactly? If you just want to highlight the regions of the image that are at that middle grey value you can create your own false colour LUT here: LUTCalc
For what its worth, i share my experience on this topic. I have used sigmoid since i upgraded to 4.6. For me the new workflow offers better control over saturation of colours (especially greens). When boosting saturation in colour balance RGB i can push it further before the colours look unnatural. And i tend to use less modules and instances with sigmoid. Take it with a grain of salt what i just wrote. Its just what i observed while editing and is not based on a mathematical or scientific background. I hope it helps anyway.
The green areas represent middle grey, the peachy colour on the forehead is 1 EV above middle grey (common fair skin tone target for cinematographers) and the blue is “just above black clip” which can be set as many EV’s below middle grey as you like. Not sure if it picks a sensible default when you pick sRGB gamma or not. You have purple for hard shadow clipping and colours for near clipping and clipped in the highlights too - or you can turn them all off.
It’s probably no easier than applying the preset you use currently, but perhaps more legible? Depends how used to it you are.
Edit: you can make them for a linear scene referred gamma too if you like and move the lut3D module where ever you like accordingly if you’re trying to find middle grey at some other point in your workflow. I just use this sometime as a bit of a sanity check before exporting because my workspace is not really geared towards ultra accurate colour work. If I don’t my images are often too dark on phone displays for example.
Yes, it is. A tool that just replaces the hue with false color based on EV in an image (not in a module, but a simple button like gamut, focus etc) would be a great addition to DT.
I will see if I can add something to the discussion.
Here is a short version of how I see it:
Filmic lets you decide a min and max brightness to fit on your display and then finds a s-curve (with your input on the look tab) that is smooth that fulfills that.
sigmoid lets you define how your greys should look (how contrasty they are) and then gives you a smooth rolloff to black and white. It’s mathematically defined for -inf EV or + inf EV but it will be a point were there is no contrast left and that will look like pure white for example.
The maybe more noticable difference between the two is that filmic uses “preserve color” per default which is the similar to the RGB ratio option in sigmoid.
And for the middle grey discussion.
0 EV = 0.1845
This is inspired by the ACES standard used in the cinema industry and its also roughly equal to Lab 50 %
That means that
-1 EV = 2^-1 * 0.1845 = 0.5 * 0.1845
1 EV = 2^1 * 0.1845 = 2 * 0.1845
2 EV = 2^2 * 0.1845 = 4 * 0.1845
And so forth.
A simple trick for “finding” middle grey is to crank up the contrast in sigmoid to 10 and then adjust exposure. But it’s not like a face should be middle grey for example so it’s better to trust your gut in most cases anyway.
It is interesting to see what happens if one loads this into darktable, boosts exposure by say 3 EV, and applies different sigmoid and filmic settings.
