vkdt devel diary

…could you specify this a bit more? you should get a *.nprof.jpg in the bin/ directory where you executed it, as well as a data/nprof/*.nprof file with the camera name and iso value in the place of the asterisk. the automatic noise profile detection only works if you have exiv2 support enabled (something VKDT_USE_EXIV2=1 in config.mk, see config.mk.defaults for explanation).

  1. I have no config.mk, I only have a config.mk.defaults
  2. I have compiled it with exiv2 support. However, in config.mk.defaults, before VKDT_USE_EXIV2=1 there is a #
  3. in /bin I have a --0.nprof.jpg, but there are no files in /bin/data/nprof/

okay, so you don’t have exiv2 support compiled in. citing https://github.com/hanatos/vkdt/blob/master/readme.md, " …can be set in config.mk. if you don’t have that file yet, you can copy it from config.mk.defaults." you probably want to do that and remove the comment sign # in front of the exiv2 line and maybe a few others. editing the defaults file directly is not a good idea, it’ll be overwritten by potential git pulls in the future. also for packaging, i can just provide config.mk entirely, makes for an easy patch.

Ok so I will try that. So I actually need to recompile vkdt? And probably I should also set RAWSPEED_PACKAGE_BUILD=1

Ah, ok, so denoise works now. Let’s try again color management.

well for rpm based distros I have a base package already but last time we discussed this, you didnt want the automatic setup yet :slight_smile: just find me on irc.

main thing is you need an account on OBS so you can set up the trigger or ask Roman to do it. darktable goes via his OBS account.

color management works finally, too. However, your instructions were not quite clear. I did this: I ran read-icc.py /home/anna/benq.icc. Then I created a file display.HDMI-0 and pasted the content of display.profile into it.
Well, thanks for the help.

right i probably need to setup some better scripting for this which at least asks xrandr and gives you the option to pick the screen or so.

these are one-time configuration things, they are not streamlined well (to say the least).

thanks. so i guess i can just create and account and will bother you on irc for all the rest.

the package build options are just for distribution of binaries, i.e. if you build on one machine but will execute on another (you may not have the latest avx512 whatnot instructions there). compilation will only enable generic x86-64 optimisations in this case.

…in other news, i’ll push editable hotkey support in a while:

I recall numpad is different from numrow, at least in normal dt…

that is correct. working on a laptop probably deceived me into disregarding the numpad completely for now. also the layout here is not my code but an imgui hotkey hack by someone else (see source for credit; who also probably works on a laptop).

this is now done:


still not entirely happy with it (hot pixel, a bit of contrast loss), but i’ll be able to work with this for a while.

i know a few bugs and maybe i want to make configurable hotkeys store/load to/from config file before a release, but that’s it.

The less defined edges tend to blur much more, except for the hot pixel, which bleeeds in one direction. Seeing the waveform becoming more defined after the denoising is neat. Edit The fringing worsens as well, which I suppose will be addressed eventually along with the hot pixel.

Thats a great image to assess noise…would you be able to share it?? No big deal if its an issue…The text on the gloves and the wood grain are nice markers for how well details hold up

… this particular one also has a face (if showing the whole frame) so i’d rather not share it. let me look through the collection whether there’s a similar one without a person.

1 Like

right, i think these are good observations. some of that i may be able to tweak away. it’s a bit messy because this is pre-demosaic, so the first step of denoising works on different data than the rest, and thresholds need to be adapted. i’m guessing the non-isotropic blur around the hotpixel is caused by the alignment of the bayer block too.

…still trying to find out how exactly the sophisticated toolchain in debian works, but if you feel adventurous (or want to give feedback):

here’s an initial debian package [3.0 MB]

it’s x86-64 and probably depends on some unstable/sid packages. will figure out how to obs it later.

do you have the dsc file + debian/ subdir as well? then I can add that to the OBS