Pride orange smile, sharpness and color challenge

////
retouch_pole_black_dot_IMGP2775.DNG.xmp added with the pole and black naval dot removed, no other modules edited.

As used in #40 (pole only), not that clean, works OK on a burned out sky.

Pole is also retouched in #45 by @dqpcoxeas .

Black dot removed by @Popanz #13.

////

This is from my first Sigma 105mm F2.8 session on a 2nd hand Pentax K-50 that slowly developed the common aperture block. I didn’t realize that and ended up shooting most of the session too slow and too noisy.

This one would be great to rescue.


The xmp is from my first attempt. The masking halo doesn’t really work but you may get what I’m going for to emphasize biggest smile on the block.

I’m considering pulling her out of this scene entirely or blurring most of her fellows out.

Recently I discovered https://upscayl.org and found its attempt really good, sharpening in particular. I’ve never really used DT sharpening.

Black dot on the the guy to the right is probably from dirty K-50 sensor.

Here’s some context

-and there’s a few more examples (some upscaled) from the session at https://photos.app.goo.gl/wfQrTAiF3R9jYage9

This work is licensed under CC BY-NC-SA 4.0

IMGP2775.DNG (14.1 MB)
IMGP2775.DNG.xmp (28.8 KB)
IMGP2775-BW.DNG.xmp (26.9 KB)
retouch_pole_black_dot_IMGP2775.DNG.xmp (11.8 KB)

1 Like


focus peaking in Darktable, looks like it went for the wall some 10-15 meters behind them

If you want to play with neutral networks for sharpening, I’d recommend chaiNNer, as it gives you more control.

Models (one can find more on GitHub):

This guy has published several deblurring models:

You can find tutorials on YouTube.

1 Like

thanks a lot, I’ll have a look and did notice chaiNNer (and Cupscale) already from Philip Hofmann’s newbie friendly intro & comparison site Interactive Visual Comparison of Upscaling Models | Interactive Visual Comparison of Upscaling Models. I’m not sure if either will run on this 2GB VRAM (Nvidia) retro laptop.

Do you know how Upscayl works? I can’t parse these docs yet - Guide · upscayl/upscayl Wiki · GitHub - it’s definitely using the Nvidia GPU and didn’t ask for pytorch at launch like Cupscale.

Philip mentioned somewhere that he chose Gigapixel in the end to skip geeking totally out on all parameters known to man and machine :slight_smile:

My version…

IMGP2775.DNG.xmp (24,4 KB)

1 Like

IMGP2775.DNG.xmp (13.3 KB)

1 Like


IMGP2775.DNG.xmp (13.2 KB)
chainner.chn.txt (2.9 KB)
(rename .chn.txt to .chn)

1 Like

Almost too sharp, wow :slight_smile:

Do you run chaiNNer sharpening before or after DT?

Darktable came first. Maybe I should have skipped the filtering, just export after basic development, then use chaiNNer, and then process further.
Here it is: only filmic etc. in darktable, then chaiNNer, then darktable with diffuse or sharpen with the local contrast preset + the local contrast module with default settings.

It’d probably be good to mask the results and/or experiment with models, as there are weird artefacts, such as:
image
Or how the hair is only sharpened inside a block:
image

@dqpcoxeas @age thanks! Very curious to dig through your sidecar and learn new DT tricks

I remain tempted to try and isolate her, remove the pole behind her right hand and the concrete chunk of a building as well. Blue shirt guy breaks the dress code and is out of there soon.

Didn’t antipate that at all before this round, thanks for the inspiration.

@kofa I’ll leave the tech notes for now and another thread, your feedback and example edit is much appreciated

1 Like

Bits and pieces (from kofa’s entry)

AI codeformer robots revealed a definite Pentax fixation. Not sure if there’s space for two portraits within this busy frame and Jubilant Stretch doesn’t need that much sharpness to stay happy. Mono seems to balance her joy better.





IMGP2775.DNG.xmp (8.0 KB)

1 Like

puhhh, that’s a hard one. I seldomly felt such helpless on a pic in Play Raw.
I’m still not convinced with my render. Anyway here it is:


IMGP2775.DNG.xmp (27,4 KB)

2 Likes

Your effort is much appreciated!

I really like your lighting and tone curves on main subject, like @dqpcoxeas above. I also prefer the crop he and @age did to include the skirt.

Her face sharpening is too strong IMO as is the outline on the left arm. But great to see what DT is capable of on its own.


Some quick Blur Background - Pixelcut mock-ups, based on your entry. They may work as a trio portrait, considering @kofa 's sharpening work on all three faces.

Hope it’s ok not to play raw (yet) during play raw!


1 Like

Another try. More colourful this time.


IMGP2775.DNG.xmp (22,1 KB)

3 Likes

ART, Tone equalizer + contrast + CTL script Film density.

IMGP2775.jpg.out.arp (11.4 KB)

3 Likes

Love it! Are you working on kofa’s jpg to achieve such face sharpness?

It’s now becoming obvious that the grey-blue-ish buildings I caught them in front of is a major part of the color clash and challenge. Yours and @age 's seems to mix up best on my display.

Maybe the Sigma 105 met a soft limit at this scene or K-50 is damaged. I ran some quick DT styles to for reference.

I:
film emulation_ Fuji xtrans ii classic chrome v1
Oly_OM-D EM5 II_Monotone_green

II:
Kodachrome 64

III:
Oly_OM-D EM5 II_Portrait

IV:
_Pastel

five_DT_styles_from_orange_pride_play_raw_example.zip (9.4 KB)

1 Like

wonderful, that orange red palette is exactly what I’m going for, skin tones blends very good as well, thanks!

No, it’s darktable alone. The XMP is attached, feel free to examine it.

1 Like


IMGP2775.DNG.xmp (19.4 KB)

4 Likes