ART (the software) news

training on local editing

I was doing some training on the photo from [Play Raw] cumulus beach

I made this correction DSC_4165.nef.arp (24.7 KB)

And I observe what I think is a bug.

  • open “snapshot1”. there are 3 color corrections defined.
  • select the first color correction and modify correction parameters to whatever you want
  • click on the second correction, the third, come back to the second. surprise, the parameters are modified accordingly to the first correction.

Can you confirm it is a bug? Then I will open a ticket.

Yes it occurs at export and it is fixed.
Thanks a lot Alberto

confirmed. I should have fixed it about one hour ago… can you please test?

Thank you. Working ok

ART-W64NightlyBuilds/ART_master_0.3-45-g9f62dad3c_W64_SSE4_191225.7z uploaded at
https://keybase.pub/gaaned92/ART-W64NightlyBuilds

@agriggio @sguyader A new french translation file. It concerns line 2003 and 2006. Slightly shorten check button text to allow a normal shrinking of the right panel.
I updated this post to take into account the new lines 813, 1970, 1971 and 1975 for Log encoding optimisationFrancais.txt (150.7 KB)

I stumbled upon this problem

I tried to use the same minimum processing in RT and ART

  • open it in RT 5.7-374-g898527732 with minimum processing: DSZ_4168.NEF.pp3 (12.1 KB)

  • open same image in ART 0.3-46-gb9dc5b5b6 with minimum processing:DSZ_4168.NEF.arp (65.2 KB)

I notice the following facts:

  • the two log-log histograms are very different. In the ART photos there are clipped highlights and shadows in contrast with RT image. I don’t know what is correct.

  • in ART I am unable to get back the shadows and highlights without destrying the image.
    I am surely missing something. Thank you for help.

Hi,

Well, the pipelines are different, so this is not surprising.

The ART histogram will show a colored rectangle at the border of the histogram if there are any clipped pixels in the image. In RT, the logic for showing the rectangles is different, and I actually cannot quite understand it, sorry.

I don’t know what is correct.

You can use the inspector in “raw clipped pixels” mode to confirm that there are clipped pixels in the raw file:

I’m not sure I understand what you mean here. Can you be more specific? If you you mean that you want to get rid of the clipping indicators in the histogram, then it’s not surprising that it requires a lot of work and the result is not nice, since those pixels are clipped already in the raw.

Hope this helps!

Yes, definitely. In rawproc, I just set the whitepoint at the minimum of the channel maximums and pray a contrition to @aurelienpierre:smile:

I’ve now compiled both RT and dt current, and in the new year will mess with the respective reconstruction capabilities, particularly RT so I can understand how to use the librtprocess recovery function. Thing is, some scenes are just challenging this way, and for most cameras putting those highlights in definition will make dealing with the shadows even more challenging, no matter the software.

Looking forward to our visit to the same cabin next year, so I can set up an equivalent scene, expose it ETTR, and see how the shadows fare from the Z6 sensor…

1 Like

@agriggio I updated the french translation file, which include all the improvements you bring recently (new lines 821, 1983, 1984)Francais.txt (150.8 KB)

  • They should not be that different as I only use amaze demosaicoing, white balance with same parameters, color management with same parameters, all exposure parameters set to 0 (no clip out of gamut for RT).
    So there should be some hidden processing or a bug in one SW to explain the difference: ART mid and high luminance pixels are more exposed than in RT.
  • You are right. if in RT I click on clipped highlight indication and clipped shadow indication, I can see clipped highlght and clipped shadow zones. Nevertheless there is no indication in the histogram.
  • I do see the raw clipped highlights. but for clipped shadow I can hardly see them. I only see some colored pixels in the shadow zone.

  • clipped shadows indication zone : when I hover on the clipped shadows zone, the RGB values seem always greater than 0. I imagine there are only a few clipped pixels in the zone.

I tried in ART to get the same non clipped histogram as in RT. That was indeed a complete failure.

hi,

those are your clipped (clamped) pixels. the colour indicates which channel is below the black level

ART and RT apply camera input profiles in a (slightly) different way, so that might be it. I’ll take a closer look when I have the chance

Just curious is there an explanation of how the pipeline for ART differs from RT and why the changes were made? (Just a brief overview if it’s not to complicated to explain?)

It is indeed a bit complicated to explain. There are different reasons:

  • to have a clear separation between tools operating in “scene referred” space and tools operating in “display referred” space, similarly to what is happening e.g. with darktable

  • to have a more modular organisation of the code

  • to fix (what I perceived to be) some inconsistencies in the way some tools were applied

  • finally, it is likely that some changes were purely accidental

Just to be clear, though, I think there’s nothing wrong in the way things are done in RT. I just had a different opinion about some things and wanted to try some alternatives. Not being constrained by backwards compatibility (which is very important for RT) made experimenting quite easy.

5 Likes

That sounds good, thanks for the explanation.

@agriggio
I updated the line 1124 (sharpening tool). Do not hesitate to ask me for a verification of the french language file before you decided to release the 1.0 version of ART.
Serge MoreauFrancais.txt (150.8 KB)

thanks a lot @srgmro for your help! I am finalising a couple more things and then I plan to publish a first release candidate. I don’t expect changes in the language files but I’m not sure yet…

@agriggio
A New french translation file, according to the modification introduce by @sguyader in the default language file. I had only to actualize the line 862 others were already OK.Francais.txt (150.5 KB)

Yes I had made changes already in the French file, I missed one occurrence. Thanks Serge.

I just want to mention this, so it may help someone who has similar issue:
I noticed that after upgrading from 0.2 to 0.3 via builds provided here for windows, the exported image when opened in other sw than ART was very midle grey-ish, low contrast. Same image exported with linux version self compiled using same sidecar file was ok after opening in same sw.
Removig local app data on windows fixed the issue and now both exports looks identical.