ssimulacra2 VS Primitive Pictures

https://github.com/cloudinary/ssimulacra2?tab=readme-ov-file... shows a higher correlation with human responses across 4 different datasets and correlation metrics for 1

Very well written. As others have suggested, you should do a followup using modern perceptual quality metrics like VMAF and ssimulacra2.

I was thinking about that too. Can't answer the question, but I did come across this just the other day: https://github.com/cloudinary/ssimulacra2 Supposedly good for comparing image similarity. Might depend on your use-case, I think it's geared towards image quality moreso than similar photos.

That's the kind of feedback I was looking for, thanks. I usually go by eye, but at 14:1, I was expecting a blurry artefact-ridden mess, and it just didn't look all that bad at first blush.... I'll definitely check out ssimulacra2, thanks for the suggestion. I also need to sit down and do some proper pixel-peeping on a better screen... I think my monitor falls into the 'HD-aren't' class of HDR displays, so I'm likely missing detail somewhere. My purpose is archival, I want to preserve quality but not needlessly waste drive space on less efficient codecs. I'm not concerned by the resultant bit rate, so long as it has improved compression efficiency over the original.

The fork to build - rust-av/ssimulacra2_bin | original release (works on images only) with a lot of info about the metric - cloudinary/ssimulacra2

TIL about https://github.com/fogleman/primitive from https://en.m.wikipedia.org/wiki/Comparison_of_raster-to-vect...

Looks great! I've been thinking of having my own attempt at something like this ever since I saw a similar idea at primitive.lol

Looks like it's made with https://github.com/fogleman/primitive

If you're looking for a more artistic arrangement of coloured blocks, and you've got some skills (or a Mac), there is this one application that's fun to play with. But I'm sure there are other ways.

Some examples off the top of my head (not game-dev specific) are https://www.michaelfogleman.com/ (creator of https://primitive.lol/), Marco Arment (https://marco.org/ - Overcast), or Pieter Levels (sadly as I write this I found that his blog seems to have been taken down and just redirects to his twitter https://twitter.com/levelsio?ref=levels.io - but still has a laundry list of projects old and new in his twitter bio)

Any ideas on how I could simply cross-compile a non-module GitHub repository to three different architectures with predictable filenames? I wouldn't mind setting up a go.mod file, the problem is I would like to compile the executable in the github.com/fogleman/primitive repository, not my own code.

Consider an SVG from [Primitive](https://github.com/fogleman/primitive) and a mostly-transparent overlay containing the diff against the original image. Tuning Primitive is not entirely deterministic (like your algorithm) and I think a human interface for the front-end would allow a few knobs to erase parts of the diff in order to shrink the detail imperceptively. Seems like a good opportunity for a contest :)

I think stroke-based painting methods like these are neat: https://github.com/jiupinjia/stylized-neural-painting https://github.com/fogleman/primitive