I have some photo’s with sensor spots. At what stage in the pipeline should I retouche them. As early as possible, before any color/tone corrections, or late in the pipeline after all corrections?
Welcome to the forum!
If performance might be an issue - I would recommend doing it first. Since the retouch module is quite early in the pixelpipe, it will take longer to re-render the image if you do it last.
Otherwise, it makes sense to do it as one of the last steps, as edits can make certain spots more visible than they are early in the process (speaking from experience).
I recommend cleaning the sensor, if possible, the wet cleaning kits are not expensive and if you do it right, the results are worth it. It is not very hard, although scary for the first time.
Please understand that the order of operations (how they are executed) is not the same in which you perform the editing steps.
Please read:
- darktable 4.9 user manual - the pixelpipe & module order
- darktable 4.9 user manual - the history stack
So, unless you change the pipeline (execution) order (if you are just beginning with darktable, do not do that!), it does not matter (does not influence the outcome) if you remove spots early on or late.
The performance (how quickly the editor reacts) is influenced by editing order, as @Vente mentioned above.
From the point of view of getting the fastests editor performance, what do you reckon is the best editing order?
Apply heavy modules late. denoise (profiled) and diffuse or sharpen are probably the heaviest ones, especially if you do not have a GPU. If possible, get one (with as much memory as possible), it boosts performance by a lot (depending on your CPU and the GPU, it could be 2 to 10 times faster). With my modest, old Nvidia 1060, denoising is not an issue, and diffuse or sharpen is also no big deal on my 4K display, though I do have to wait a bit if I enable the full-resolution editor preview.
RTX 4060 8GB should be perfect, right? (24 Mpx images)
I 100% agree with this comment, but on my personal computer I apply denoising and initial sharpening as one of the first steps because my computer is faster enough with its dedicated GPU. However, if you find yourself waiting for these heavy modules then apply them last as suggested.
I tend to do the sensor spots near the end as they don’t influence my editing decisions and just need to be taken care of before finishing off the edit.
Thanks for the advice @kofa and @Terry. (I’m running darktable on Debian 12 installed on a very elderly 27" iMac, so no real option for adding a chunky GPU.)
Thanks all,
I’m still learning a lot
I’m using two older computers right now. A 2014 macbook pro 16GB, which becomes unworkable with the diffuse&sharpen module (lr was slow, but not unworkable)
I also hijacked my son’s old 2017 game laptop which has also 16GB and a nvidia gfx1060. That one is responsive.
About cleaning the sensor: of course. The dust appeared during a desert trip when I did not have the cleaning kit with me.
I also agree, apply the “heavy” modules at the end of the editing process. You should also consider, how important the respective modul is for your editing process. Changes in color or brightness in most cases also influence other editing decisions. So you should do them early. Moderate noise reduction on the other hand will not change you overall editing process. This can be done in the end.
Same situation here, at least with my old, relatively ‘low-resolution’ (10 MPx, 16 MPx) cameras. When my display was ‘only’ Full HD, it was even faster.
I tested out the weather sealing of my Canon R7 in a sandstorm in the Sahara desert. It actually survived, but I don’t recommend that test to anyone. It’s like putting your camera in sandblaster.
Keep in mind that you can also disable modules while removing the spots, and then re-enable them once you’ve finished (they will keep their settings for the current image).
It also may be useful to use several copies of the retouch module if there are many spots to remove (you could even disable the earlier copies while working with the latest, but keeping track of what’s done may be more complicated)
My go-to solution to that kind of environment is something like a transparent plastic bag that encloses the camera, fixed around the lens hood with tape, and a clear filter you will dispose of later on, and consider your lens hood disposable too. If you can get your hands on 4–6 mil plastic bags you can even turn the dials through it and it will last quite a bit, but take spares.
I would put the module right after the modules that just deal with the sensor as is, ie raw CA correction, demosaic, profiled denoise etc, and importantly before lens correction. If dirt is on the sensor, this may allow you to figure out the spots just once, or a few times you changed lenses if they kept accumulating, and copy over the correction to all images affected.