|
portrait
[nfo]
|
||||||||
|---|---|---|---|---|---|---|---|---|
|
|
|||||||
|
popularity : 56% |
|||||||
alltime top: #10643 |
|
|||||||
|
||||||||
added on the 2018-01-15 23:00:41 by Witek  |
||||||||
popularity helper
comments
Really nice! And esp. nice to see somebody experimenting with new techniques instead of going the "usual way".
rulez added on the 2018-01-15 23:29:10 by v3nom 
Impressive!
Looks quite nice! Interesting idea with the voronoi bits.
Cool thing again.
What did you use as the fitness function in your genetic algorithm? PSNR or something more specialized?
What did you use as the fitness function in your genetic algorithm? PSNR or something more specialized?
Impressive technique!
Quote:
What hat did you use as the fitness function in your genetic algorithm? PSNR or something more specialized?
Yup, just PSNR. I was experimenting with SSIM, but it was too slow to compute.
very very nice!
Wow!
^
↑
Works really well!
norm
this is amazing! well done!
Witek: I tried to run this at work and it got snagged by the AV :) . For IT's peace of mind, which compressor did you use? I presume Crinkler, but I'd like to be able to tell them for sure. Thanks!
saga: nice oldschool uparrow :)
saga: nice oldschool uparrow :)
Wow!
Looks much better on screenshot than in full screen, but still rather impressive. :)
Witek: to my understanding PSNR and SSIM in "real world" use cases like this aren't that much different. Both of them are actually somewhat weak since they don't correlate that well with human visual acuity. Seems to me like the biggest shortcoming might be there, so was just wondering how much better it could look with a more specialized metric.
Awesome! Did your fitness function use a perceptual colourspace?
Quote:
Witek: I tried to run this at work and it got snagged by the AV :) . For IT's peace of mind, which compressor did you use? I presume Crinkler, but I'd like to be able to tell them for sure. Thanks!
Yes, it's just Crinkler. No worries :)
Quote:
Witek: to my understanding PSNR and SSIM in "real world" use cases like this aren't that much different. Both of them are actually somewhat weak since they don't correlate that well with human visual acuity. Seems to me like the biggest shortcoming might be there, so was just wondering how much better it could look with a more specialized metric.
It depends. Generally, SSIM should give more perceptual-like results and outperforms simple PSNR in most cases. Maybe if I have moved the calculations to the GPU and it would be 2 orders of magnitude faster than CPU version (which is most likely) I could experiment with different quality scores.
Quote:
Did your fitness function use a perceptual colourspace?
For this prod, I've just used monochromatic value for calculating MSE (calculated as Y=R+2G+B).
Amazing.
what they said
Not that you would notice a difference with current lossy formats dealing with it in a second, but still the tech is impressive.
cool
Nice!
Now I know, this reminds me of Sofia Coppola.
Very nice.
submit changes
if this prod is a fake, some info is false or the download link is broken,
do not post about it in the comments, it will get lost.
instead, click here !