|
yurume
|
2023-02-03 07:31:18
|
yeah, that sounds like a good idea (for who wants watermarking)
|
|
2023-02-03 07:32:35
|
it can be "removed" by detecting watermarks (impossible to prevent) and reconstructing details disregarding the original image behind the watermarks, but at this point you can probably generate much of the image with that model anyway
|
|
|
_wb_
|
2023-02-03 08:21:26
|
The idea is that e.g. a stock photography website would apply the effect on all its images. For any given image, the watermark is only applied once, not multiple times. But since the watermark is subtly different for every image on the website, it's harder to let AI auto-remove the watermark.
|
|
2023-02-03 08:22:32
|
Something else: I'll likely be at FOSDEM tomorrow. Anyone else happens to go there?
|
|
|
|
JendaLinda
|
2023-02-03 09:50:29
|
It surprised me that JPEG at quality zero can't even represent pure white and pure black.
|
|
|
diskorduser
|
2023-02-04 05:36:34
|
https://youtu.be/58k28fQiv9U
|
|
|
|
JendaLinda
|
2023-02-04 10:50:41
|
That's why I don't use laptops anymore. I had so much trouble with laptops. I hate them. I've built myself a desktop PC from used components and I'm extremely satisfied.
|
|
|
|
_wb_
|
|
Traneptora
|
|
_wb_
|
|
2023-02-04 01:45:17
|
is this Free Beer?
|
|
2023-02-04 01:45:45
|
https://en.wikipedia.org/wiki/Free_Beer
|
|
|
|
_wb_
|
2023-02-04 02:04:36
|
Unfortunately not free, but it was a good beer
|
|
|
daniilmaks
|
2023-02-04 08:23:13
|
was it at least free as in freedom?
|
|
|
|
_wb_
|
2023-02-04 08:34:57
|
What does that mean for beer? I didn't get the recipe for brewing it, but then again I also didn't ask...
|
|
|
Nova Aurora
|
2023-02-04 08:50:23
|
The right to use it for any purpose
|
|
|
|
Traneptora
|
|
_wb_
Unfortunately not free, but it was a good beer
|
|
2023-02-05 05:49:40
|
Free Beer is an open-source beer. it's libre, not gratuis
|
|
|
|
yurume
|
2023-02-05 07:15:23
|
as in https://en.wikipedia.org/wiki/OpenCola_(drink)
|
|
|
DZgas Ж
|
|
|
yurume
|
2023-02-05 10:59:33
|
some use WHT, don't they
|
|
|
|
DZgas Ж
|
|
yurume
some use WHT, don't they
|
|
2023-02-05 11:09:09
|
some is who? I only know about Jpeg2000
|
|
2023-02-05 11:09:42
|
And couple of completely unknown indie codecs
|
|
|
|
Traneptora
|
2023-02-05 11:25:13
|
Jpeg XL lossy modular based on haar wavelets
|
|
|
|
DZgas Ж
|
|
Traneptora
Jpeg XL lossy modular based on haar wavelets
|
|
2023-02-05 11:26:12
|
Jpeg XL contains 2 independent codecs
|
|
2023-02-05 11:27:01
|
and the codec could not do without DCT
|
|
|
|
_wb_
|
2023-02-05 12:30:16
|
DCT is just a very good idea for lossy compression, people have tried other stuff like fractal compression and wavelets but nothing really beats good old DCT
|
|
|
|
DZgas Ж
|
|
_wb_
DCT is just a very good idea for lossy compression, people have tried other stuff like fractal compression and wavelets but nothing really beats good old DCT
|
|
2023-02-05 07:55:19
|
it is really strange that have invented with so few ways to transform information for lossy compress
|
|
2023-02-05 08:06:03
|
oh well https://en.wikipedia.org/wiki/List_of_Fourier-related_transforms
|
|
|
Demez
|
2023-02-07 08:04:40
|
<@1028567873007927297>
https://discord.com/channels/794206087879852103/804324493420920833/1071760452402561094
I mean, I think it a good use of it is a pure virtual class with the use of override for stuff across dlls, that's what I do
and maybe for multiple options for something, like audio decoders, or image codecs
|
|
|
|
yurume
|
2023-02-07 08:06:30
|
but that's completely unrelated to object orientation
|
|
|
|
Demez
|
2023-02-07 08:06:48
|
I mean, it uses classes and inheritance to do that though
|
|
|
|
yurume
|
2023-02-07 08:07:08
|
not even inheritance, you can't "inherit" any behavior out of it, just interface
|
|
2023-02-07 08:07:45
|
more generally speaking "OOP" is such an elusive concept that no one can't agree what does it actually mean
|
|
|
|
Demez
|
2023-02-07 08:07:57
|
I guess so lol
|
|
2023-02-07 08:08:36
|
I always deemed anything that uses a class "OOP"
|
|
|
|
yurume
|
2023-02-07 08:08:40
|
inheritance is not a strict requirement for OOP, class isn't (cf. pre-ES6 JS), encapsulation isn't (cf. Python)
|
|
2023-02-07 08:09:58
|
I generally think OOP as a very loose term to group somewhat similar looking languages back in 90s, and nothing else
|
|
|
|
Demez
|
2023-02-07 08:10:19
|
one thing i really dislike are classes that are only used as a singleton and in the same binary, like `class Input` and `Input gInput` or something
|
|
|
|
yurume
|
2023-02-07 08:11:35
|
maybe the only actually common feature across "OOP" languages is a receiver-method syntax (`foo.method(...)`)
|
|
2023-02-07 08:12:14
|
which is no small idea, because it means that a namespace for `foo` methods can differ from one for `bar` methods
|
|
|
w
|
2023-02-07 08:12:16
|
I consider oop anything that normally persists states (objects), specifically the possibility of mutation
|
|
|
|
yurume
|
2023-02-07 08:12:24
|
and this ultimately allows for polymophism
|
|
2023-02-07 08:12:52
|
there are other ingredients to make OO-style polymorphism to work, but I believe syntaxwise it is the biggest factor
|
|
|
Demiurge
|
2023-02-07 08:21:08
|
I consider OOP to be a bunch of unnecessary boilerplate and unintuitive awkward forced abstraction
|
|
|
|
yurume
|
2023-02-07 08:22:07
|
I believe even that is an useless statement because you can't rigorously define OOP anyway
|
|
|
|
Demiurge
|
2023-02-07 08:22:14
|
> “The phrase "object-oriented” means a lot of things. Half are obvious, and the other half are mistakes.“ – Paul Graham
|
|
2023-02-07 08:24:04
|
I have no interest in even going anywhere near attempting to define OOP
|
|
2023-02-07 08:24:25
|
It's basically whatever anyonewants it to mean already
|
|
|
|
yurume
|
2023-02-07 08:24:25
|
without actual definition, it is useless to argue for or against it 😛
|
|
|
|
Demiurge
|
2023-02-07 08:25:19
|
Well then my current working definition is "unnecessary boileplate and unintuitive forced abstraction" in the context of programming languages like Ruby and Java
|
|
|
|
yurume
|
2023-02-07 08:25:36
|
we should dissect OOP into individual parts and discuss them in isolation, because it makes much easier to see what's good and what's bad
|
|
|
|
Demiurge
|
2023-02-07 08:26:53
|
People should just say what their software is supposed to do in their code, in the most straightforward and simple way possible. Anything that gets in the way of that is a mistake
|
|
|
|
w
|
2023-02-07 08:26:58
|
I swear it is just objects/mutation when contrasting with non-oop
|
|
|
|
Demiurge
|
2023-02-07 08:27:03
|
And an obnoxious one
|
|
2023-02-07 08:27:57
|
Trying to force people to use abstractions in situations where there doesn't need to be one, is evil and stupid
|
|
|
|
yurume
|
2023-02-07 08:28:06
|
receiver-method syntax: not that bad, but forcing this everywhere is not a good idea either; inheritance: almost universally regarded as bad, prefer composition instead; state: tolerable if and only if isolated enough; encapsulation: generally good but you will eventually need backdoors anyway
|
|
|
Demiurge
Trying to force people to use abstractions in situations where there doesn't need to be one, is evil and stupid
|
|
2023-02-07 08:28:23
|
for example?
|
|
2023-02-07 08:29:02
|
I mean, heavy abstractions are generally bad signs, but not all abstraction is unnecessary so I want to see which one you regard as unnecessary
|
|
|
|
Demiurge
|
2023-02-07 08:30:28
|
Well like forcing this whole "object" and "class" and whatever semantics when it's not natural to how a computer works nor to how people think either, despite proponents saying "lol it's so natural for humans to think of everything as objects..." Bro, I am trying to do something really simple here, don't overcomplicate it with stupid analogies...
|
|
|
|
yurume
|
2023-02-07 08:31:38
|
I guess then you were referring to object-*only* languages, and I agree to you as I've said "forcing this everywhere is not a good idea either" above 😉
|
|
|
|
Demiurge
|
2023-02-07 08:32:16
|
Well tell that to Java and Ruby and all these other object-oriented hypetrain languages
|
|
|
|
yurume
|
2023-02-07 08:32:44
|
but I wonder about your opinion on not-actually-object-only languages, say, Python
|
|
|
|
w
|
2023-02-07 08:32:53
|
yeah changing a variable in a for loop is stupid
|
|
2023-02-07 08:33:00
|
i love elixir
|
|
|
|
Demiurge
|
2023-02-07 08:34:06
|
Python seems like a far more reasonable language, yeah.
|
|
2023-02-07 08:34:21
|
Because it doesn't force its object oriented shlong down your throat
|
|
|
|
w
|
2023-02-07 08:35:06
|
Tbf java doesn't either
|
|
2023-02-07 08:35:13
|
You can use classes statically as modules
|
|
2023-02-07 08:35:26
|
For functional
|
|
|
|
Demiurge
|
2023-02-07 08:36:42
|
changing a variable in a for loop is a little bit stupid, compared to some way of some "x times loop" statement, but for a simple language lacking that, creating a counter variable is the most logical and obvious way to go
|
|
|
|
yurume
|
2023-02-07 08:36:45
|
mainly because Java uses the same unit (class) for both function namespace and compilation unit
|
|
|
|
w
|
2023-02-07 08:37:04
|
The problem is "variable"
|
|
2023-02-07 08:37:12
|
Values shouldnt change
|
|
|
|
yurume
|
2023-02-07 08:37:14
|
in this sense, Java has a reason to not fix this; changing this will break its compilation model
|
|
2023-02-07 08:37:26
|
C#, on the other hand, is a strange mess
|
|
|
|
Demiurge
|
2023-02-07 08:37:49
|
a counter does change as it counts up or down
|
|
|
|
w
|
2023-02-07 08:38:01
|
but it doesn't need to be a variable
|
|
|
|
yurume
|
2023-02-07 08:38:05
|
C#'s unit of compilation is an assembly and not a class, so I believe free functions can exist without much headache
|
|
|
|
w
|
2023-02-07 08:38:12
|
It can be a new state in every iteration
|
|
2023-02-07 08:38:33
|
which is one of the cores of functional programming
|
|
|
|
Demiurge
|
2023-02-07 08:38:48
|
parallel programming?
|
|
|
|
yurume
|
2023-02-07 08:39:12
|
on FP: like OOP, it now shows a sign of overenthusiastic adoption I believe
|
|
2023-02-07 08:39:52
|
state is hard to maintain; but sometimes it's the best abstraction over others
|
|
2023-02-07 08:40:11
|
making everything stateless is not a panacea
|
|
2023-02-07 08:40:37
|
especially when that *state* is local to some boundary
|
|
2023-02-07 08:40:56
|
I can easily agree that the global state is bad though
|
|
|
|
w
|
2023-02-07 08:41:17
|
It's annoying at first but it just makes so much sense
|
|
|
|
yurume
|
2023-02-07 08:41:58
|
again, this is not a new idea, for example Rust preaches that that "boundary" would be a thread boundary, and within the same thread you can enjoy (non-global) states without much danger
|
|
2023-02-07 08:42:49
|
I believe this approach is much more productive than making absolutely everything stateless
|
|
2023-02-07 08:43:35
|
and I'm not a FP newbie, my master's thesis involved both Scheme and Haskell, and I had a good deal with OCaml in the past
|
|
|
|
w
|
2023-02-07 08:44:04
|
I just like the idea of spawning a process that sends itself the same data to store something in memory
|
|
|
|
yurume
|
2023-02-07 08:44:38
|
is that idea of message-passing a new OOP or FP? :p
|
|
2023-02-07 08:44:57
|
in the era of microservices I feel that too is used too much
|
|
|
|
w
|
2023-02-07 08:45:04
|
just the classic state machine
|
|
|
|
yurume
|
2023-02-07 08:45:26
|
yeah, classic state machine is no different from turing machine and they are equally hard to analyze
|
|
|
|
w
|
2023-02-07 08:45:41
|
loop that does nothing
|
|
|
|
DZgas Ж
|
2023-02-08 02:43:18
|
New neural pixelization is wow
|
|
2023-02-08 02:44:20
|
Original
Point
Spline
Neural
|
|
2023-02-08 02:44:56
|
https://github.com/AUTOMATIC1111/stable-diffusion-webui-pixelization
|
|
|
improver
|
2023-02-08 04:45:31
|
point stuff feels like it preserves color tones of original better
|
|
2023-02-08 04:46:03
|
neural is like reinterpretation of what equivalent good-looking pixel art would be like
|
|
2023-02-08 04:46:55
|
i wonder if this could be applied to generate better-looking ansi and irc art
|
|
|
|
diskorduser
|
2023-02-08 04:57:38
|
Is there any good noise reduction ai which removes noise with the help of multiple images of the same subject?
|
|
|
|
DZgas Ж
|
|
diskorduser
Is there any good noise reduction ai which removes noise with the help of multiple images of the same subject?
|
|
2023-02-08 05:04:38
|
This is the first time I've heard of such a thing... the usual good noise reduction neural networks work well only on video
|
|
|
|
diskorduser
|
|
|
DZgas Ж
|
|
improver
neural is like reinterpretation of what equivalent good-looking pixel art would be like
|
|
2023-02-08 05:07:36
|
this is an important element of pixel graphics, it is wise to limit the number of colors so that image is like pixil art
|
|
2023-02-08 05:08:11
|
|
|
|
|
JendaLinda
|
2023-02-08 05:55:21
|
Looks kinda like VGA graphics. Real pixel art would make use of ordered dithering as well.
|
|
|
|
w
|
2023-02-08 10:57:44
|
i dont think you need an ai for multiple images of the same subject
|
|
|
bonnibel
|
2023-02-11 01:06:07
|
https://twitter.com/AlexisCompBio/status/1623908738716532737
|
|
|
|
daniilmaks
|
2023-02-11 01:22:48
|
`// what the fuck?`
|
|
|
|
yurume
|
2023-02-11 02:17:58
|
turns out that the quality score contains the number of swear words as its component /jk
|
|
|
|
_wb_
|
2023-02-11 05:06:30
|
Is that distribution on the right giving us the finger?
|
|
2023-02-11 05:07:05
|
(not the middle finger though, it seems)
|
|
|
|
yurume
|
2023-02-11 06:31:41
|
that anomaly was actually the reason I joked as above, it certainly looks like that # swear words is actually a part of the quality score equation, but it was somehow capped so that we see a weird bump.
|
|
|
spider-mario
|
2023-02-11 08:18:51
|
someone in the thread mentioned a mixture of two gaussians, I’d be interested in giving a try to analysing it as such if the raw data is available somewhere
|
|
2023-02-11 08:21:31
|
eep, frequentist statistics
|
|
2023-02-12 08:50:27
|
https://www.washingtonpost.com/news/animalia/wp/2017/06/08/bear-breaks-into-house-plays-the-piano-but-not-very-well/
|
|
|
|
Demez
|
2023-02-12 09:43:53
|
has anyone tried to put jxl's XYB ICC profile into a lossless webp?
|
|
2023-02-12 09:44:40
|
actually nvm I just read this lol
https://discord.com/channels/794206087879852103/794206170445119489/1073638484268556330
|
|
|
|
spider-mario
|
2023-02-12 09:58:28
|
it will not work for video (which does not use ICC) but it should work for lossless webp, shouldn’t it?
|
|
|
|
_wb_
|
2023-02-12 11:22:19
|
For lossless, xyb is not useful imo
|
|
2023-02-12 11:23:17
|
And for lossy, webp has obligatory 4:2:0 and tv range yuv...
|
|
|
|
DZgas Ж
|
2023-02-12 04:25:24
|
lossless xyb 😂
|
|
|
zamfofex
|
2023-02-14 10:41:31
|
I suppose there is such a font readily available in my distro, so I guess I could install it. 🤔 Though on the other hand, it doesn’t cause me inconvenience frequently enough to miss it.
|
|
|
|
w
|
2023-02-14 10:42:26
|
ok guys from here on only type in cjk
|
|
|
yoochan
|
|
|
yurume
|
2023-02-14 12:42:20
|
거야 어려운 일이 아니죠
|
|
|
|
DZgas Ж
|
2023-02-16 10:31:45
|
https://cdn.discordapp.com/attachments/677266163994198020/1075905185420939374/DZgas_music_tracklist_he_aac.mp4
|
|
|
|
improver
|
2023-02-17 03:45:05
|
https://www.cafeoto.co.uk/shop/toshi-tsuchitori-spiral-arms-with-new-gamelan-ense/
|
|
2023-02-18 01:29:57
|
https://www.youtube.com/watch?v=_C3kbhC_EIY
|
|
|
|
spider-mario
|
2023-02-19 09:03:29
|
https://youtu.be/-6dCT8JL3mY
|
|
2023-02-19 09:03:33
|
there was some solid music in this game
|
|
2023-02-19 09:04:41
|
Library of Fear, The Mysterious Planet and Enchanted are excellent
|
|
2023-02-19 09:05:37
|
you wouldn’t expect a Smurfs game on the Game Boy to have a great soundtrack, and yet
|
|
2023-02-19 09:13:11
|
the previous game (same composer) as well: https://youtu.be/1pJ4Je5eiYw
|
|
2023-02-19 09:13:16
|
e.g. River Smurf or Swamp
|
|
2023-02-19 09:19:01
|
oh, how could I forget about the reggae in “Mountain”
|
|
|
|
DZgas Ж
|
2023-02-20 03:25:40
|
https://github.com/ShieldMnt/invisible-watermark
|
|
2023-02-20 03:27:29
|
half a year ago I studied this topic. it is very fragile and leaves noticeable defects in the image..... but it seems to me that half an hour ago I came up with a completely new way to save a very small amount of information on the image... the link above is useful because there are examples there - what changes and deformation of the image can affect the readability of the invicible data
|
|
2023-02-20 05:27:06
|
My very interesting methods of saving data, as it seemed to me, which works flawlessly with any transformations, even with compression. But according to my current tests, the capacity is extremely and extremely small, now I'm doing it for a black-and-white 8-bit image. in theory, more information can be stored for a full 24-bit image.
I would like to interest someone in this, because I'm a bad programmer and I write everything in python, it works miraculously for a long time
|
|
2023-02-20 05:28:26
|
the normal image. and the image after compression, in which I also recorded 2 bits of information
|
|
2023-02-20 05:30:31
|
so. what I invented. saving information in the color graph
|
|
2023-02-20 05:35:00
|
if you just take an image editor, then you can't do this there, but I, in turn, do the redistribution of colors according to the wave function - which makes the changes noticeably less because when the image is compressed, these indicators are smoothed
|
|
2023-02-20 05:36:43
|
Okay, I'll just drop it here and wait for anyone who has a normal knowledge of programming languages. my python code processes one image for the minute for 2 bit save
I would like to say that I want to make saving images primarily for lossy compression. for lossless compression, it saves hundreds of times more information.
|
|
|
|
improver
|
2023-02-20 05:40:00
|
https://www.youtube.com/watch?v=tZisr4NklN8
|
|
2023-02-20 07:18:12
|
https://github.com/psilva261/opossum
|
|
|
|
_wb_
|
|
DZgas Ж
if you just take an image editor, then you can't do this there, but I, in turn, do the redistribution of colors according to the wave function - which makes the changes noticeably less because when the image is compressed, these indicators are smoothed
|
|
2023-02-20 08:26:18
|
not sure if I understand — you're adjusting the histogram of color values in some way? (which way?) and then compress, and at decode you do the inverse histogram adjustment and get better compression results that way?
|
|
|
|
DZgas Ж
|
|
_wb_
not sure if I understand — you're adjusting the histogram of color values in some way? (which way?) and then compress, and at decode you do the inverse histogram adjustment and get better compression results that way?
|
|
2023-02-20 08:38:41
|
Compression results? No, it's not about compression, it's about storing information inside an image.
|
|
|
_wb_
not sure if I understand — you're adjusting the histogram of color values in some way? (which way?) and then compress, and at decode you do the inverse histogram adjustment and get better compression results that way?
|
|
2023-02-20 08:39:59
|
A histogram is the number of all 256 colors in an image
|
|
2023-02-20 08:40:33
|
I thought about it. That it can be used as a way to store hidden information
|
|
|
|
_wb_
|
2023-02-20 08:44:05
|
ah yes, but there are probably better ways to do steganography
|
|
2023-02-20 08:46:36
|
the histogram is not that robust if you want the hidden information to survive stuff people tend to do to images, like resize, encode with lossy compression, crop, convert between colorspaces, adjust brightness/contrast, print & scan, etc
|
|
|
|
DZgas Ж
|
|
_wb_
the histogram is not that robust if you want the hidden information to survive stuff people tend to do to images, like resize, encode with lossy compression, crop, convert between colorspaces, adjust brightness/contrast, print & scan, etc
|
|
2023-02-20 09:09:53
|
Well, that's it. I'm talking. that my way can transfer all these transformations... I've already been looking for all sorts of ways to save information on paper. and there are none. well, there are very few of them and I looked at them all. in terms of conservation, they are not even close
|
|
|
|
Nova Aurora
|
2023-02-22 03:37:54
|
<@416586441058025472> why do the ms-ttf-corefonts use bitmap fallbacks on linux?
|
|
|
|
diskorduser
|
2023-02-22 05:42:30
|
He doesn't use Linux
|
|
|
|
DZgas Ж
|
2023-02-22 07:28:40
|
<:Windows:806135372298977342>
|
|
|
Fox Wizard
|
2023-02-22 03:25:26
|
The generation loss of this meme must be very high <:kekw:808717074305122316>
|
|
2023-02-22 03:25:56
|
Got re encoded 6+ times <:KekDog:805390049033191445>
|
|
|
|
spider-mario
|
2023-02-22 05:13:30
|
how did Alberto J. González get such great sound out of the GB chip? https://youtu.be/ZXlpzbIl1Bg https://youtu.be/h32OarxPpCY https://youtu.be/b69RTU49cq8
|
|
|
|
improver
|
2023-02-23 12:26:44
|
https://www.youtube.com/watch?v=Lbfqfh3XQ10
|
|
|
|
spider-mario
|
2023-02-23 08:47:42
|
https://youtu.be/EfoHvsMOlTY
|
|
2023-02-23 08:47:50
|
“so what kind of music do you listen to?” “erm, it’s complicated…”
|
|
|
|
_wb_
|
2023-02-23 03:40:41
|
https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web
|
|
|
Mikarific
|
2023-02-26 11:13:32
|
Right... so I was invited to this discord server because I need people smarter than me to figure out what is going on with an image I was sent.
in essence, in another discord I'm in someone's phone seems to have corrupted an image and it has some very interesting properties.
|
|
2023-02-26 11:13:57
|
It looks different depending on where you view it and something causes it to look different every time it's uploaded to discord.
|
|
2023-02-26 11:14:17
|
here's the image
|
|
2023-02-26 11:14:34
|
here's the image... unchanged... just uploaded to discord again
|
|
2023-02-26 11:15:01
|
Photoshop refuses to open it.
|
|
2023-02-26 11:15:35
|
In Gimp it looks like this:
|
|
2023-02-26 11:16:32
|
In Paint it looks like this, it isn't actually solid color (besides that top part) it actually has a pattern to it (second image):
|
|
2023-02-26 11:17:26
|
In Chrome it looks like it's fully transparent black... which is VERY weird because jpeg doesn't support transparency...
|
|
2023-02-26 11:18:36
|
If anyone can figure out why this image is doing what it's doing please let me know I'm very confused and interested.
|
|
2023-02-26 11:21:32
|
sure
|
|
2023-02-26 11:21:44
|
|
|
2023-02-26 11:22:59
|
I'm wondering if the weird behavior can be reproduced in a brand new image
|
|
2023-02-26 11:24:36
|
my working theory for the discord behavior is that the image's size being so large makes discord try to downscale the image on its media proxy which causes it to look different every time but that still doesn't explain the behavior of it looking different in Chrome vs Paint vs GIMP
|
|
|
|
_wb_
|
2023-02-26 11:27:38
|
Could be a truncated jpeg maybe
|
|
|
Tirr
|
2023-02-26 11:29:04
|
basically anything can happen with a broken image
|
|
|
|
Mikarific
|
2023-02-26 11:29:42
|
im wondering if the behavior of looking different on each discord upload can be reproduced
|
|
|
|
Tirr
|
2023-02-26 11:33:23
|
inconsistent output of discord might be from randomness of internal data structures
|
|
2023-02-26 11:33:51
|
or disclosure of uninitialized memory (not likely)
|
|
|
|
Mikarific
|
2023-02-26 11:39:58
|
I have other images from this person that their phone has corrupted apparently
|
|
2023-02-26 11:40:08
|
this one doesn't embed
|
|
2023-02-26 11:40:44
|
on mobile it *tries to* but still shows up as a file
|
|
2023-02-26 11:41:26
|
in chrome it looks like a tiny white square and basically everywhere else looks like this
|
|
|
Mikarific
this one doesn't embed
|
|
2023-02-26 11:43:32
|
but this is the behavior of this image on chrome
|
|
2023-02-26 11:51:05
|
yeah...
|
|
|
|
veluca
|
|
Mikarific
in chrome it looks like a tiny white square and basically everywhere else looks like this
|
|
2023-02-26 06:45:42
|
this also seems to be a truncated image fwiw
|
|
|
sklwmp
|
2023-02-27 12:47:17
|
I'm getting some really obscure, miniscule bugs when using Thorium AVX2 versus the normal Thorium AVX:
(I guess it's better to use that build since AVX2 is more buggy, as mentioned in https://github.com/Alex313031/Thorium/issues/107):
|
|
|
sklwmp
I'm getting some really obscure, miniscule bugs when using Thorium AVX2 versus the normal Thorium AVX:
(I guess it's better to use that build since AVX2 is more buggy, as mentioned in https://github.com/Alex313031/Thorium/issues/107):
|
|
2023-02-27 12:48:45
|
I don't even know how to open a reproducible issue here, since that would require signing up for a Facebook account just to test Messenger's UI.
|
|
|
Jyrki Alakuijala
|
2023-02-27 06:54:49
|
stupid Python riddle: 'depleted boron has %d nucleons' % (0xbor 'element' + 'ChatGPT::Eval("how many nucleons depleted $element has?")')
|
|
2023-02-28 09:18:01
|
I deleted my Facebook account a few years back. I have been able to live a life of similar quality with or without a Facebook account -- I do use WhatsApp from Meta.
|
|
|
|
sklwmp
|
2023-02-28 12:03:24
|
I really wish I could say the same. Almost everybody around me has a Facebook account, it's basically impossible to survive without one for communication in this country.
|
|
2023-02-28 12:04:03
|
They even take advantage of the lack of net neutrality here by offering "Free Data" which is basically access to Facebook and Messenger, but text-only, for free, without data charges.
|
|
|
|
improver
|
2023-02-28 01:15:00
|
this kind of "free data" is the biggest scam ever
|
|
|
|
Jyrki Alakuijala
|
|
sklwmp
I really wish I could say the same. Almost everybody around me has a Facebook account, it's basically impossible to survive without one for communication in this country.
|
|
2023-02-28 03:47:48
|
Philippines?
|
|
|
|
sklwmp
|
2023-02-28 10:22:19
|
Yep. How'd ya guess?
|
|
|
|
Traneptora
|
2023-03-01 03:14:17
|
it's definitely a truncated jpeg
|
|
2023-03-01 03:17:47
|
```
JPEG XL encoder v0.9.0 60ab29d7 [AVX2]
Note: Implicit-default for JPEG is lossless-transcoding. To silence this message, set --lossless_jpeg=(1|0).
Read JPEG image with 140480 bytes.
Encoding [Container | JPEG, lossless transcode, effort: 7 | JPEG reconstruction data],
./lib/jxl/jpeg/enc_jpeg_data_reader.cc:463: JXL_FAILURE: Unexpected end of scan.
./lib/jxl/jpeg/enc_jpeg_data_reader.cc:872: JXL_FAILURE: Invalid scan.
./lib/jxl/jpeg/enc_jpeg_data.cc:311: JXL_FAILURE: Error reading JPEG
./lib/jxl/encode.cc:1561: Error during decode of input JPEG
JxlEncoderAddJPEGFrame() failed.
EncodeImageJXL() failed.
```
|
|
|
|
diskorduser
|
2023-03-01 05:34:05
|
https://play.google.com/store/apps/details?id=org.mozilla.klar
|
|
2023-03-01 05:34:38
|
Does anyone know about this browser? It looks similar to ff focus
|
|
|
190n
|
2023-03-01 07:13:46
|
https://support.mozilla.org/en-US/kb/difference-between-firefox-focus-and-firefox-klar
|
|
|
|
improver
|
2023-03-02 06:49:25
|
https://www.youtube.com/watch?v=5b9K0LWtefM
|
|
|
|
_wb_
|
2023-03-03 05:48:41
|
CMYK does not use the range 0..100 in any actual implementations I know of. That's just how Photoshop and some other software show the range in their user interfaces (they remap the internal 0..255 or 0..65535 range to 100..0 to make it easier to understand for humans)
|
|
2023-03-03 08:36:25
|
To me it looks like they basically represent RGB as 2^R 3^G 5^B instead of the usual (R<<16) + (G<<8) + B. I don't see the advantage, it is just equivalent but computationally worse.
|
|
2023-03-03 08:37:32
|
I suppose I should read the paper instead of that web page to see if there's anything actually useful there
|
|
2023-03-03 08:59:54
|
meh, how did they get that accepted in nature?
|
|
2023-03-03 09:00:45
|
Even that first sentence, what the... "Existing colour systems, such as RGB and CYMK, are all text-based" ?
|
|
2023-03-03 09:00:47
|
text-based?
|
|
2023-03-03 09:06:23
|
computationally this does not seem to be convenient at all... black is represented as 2^256 * 3^256 * 5^256, so 30^256, which is a much larger number than the largest value you need if you just use 8 bits per component
|
|
2023-03-03 09:10:26
|
also their "unification of RGB and CMYK" completely ignores the main actual difference between those two color models: CMYK has an extra component which brings an actual extra degree of freedom — you can choose to represent dark colors using less or more K, and this does make a difference in practice because reproduction is never perfect and ink and paper interact in strange ways, so e.g. doing text as pure K will look much better than doing it with a mix of CMY that is supposed to be just as black.
|
|
2023-03-03 09:16:04
|
the only maybe interesting thing is how they bucket hue+saturation combinations, but to me it looks like they remain within the HSV model, which is perceptually not very uniform
|
|
2023-03-03 09:18:02
|
anything that talks about RGB as if that's an absolute color space and not relative to the actual primaries/white point/transfer curve, and then talks about "unifying colors", seems already suspicious and simplistic to me
|
|
|
|
veluca
|
2023-03-03 09:20:14
|
that article seems... uh... I am not sure I can quite express the sentiment in an even mildly polite way... let's say a lot of pseudoscience?
|
|
|
|
_wb_
|
2023-03-03 09:20:28
|
basically the whole thing looks to me to be, at best, just a very convoluted way of saying "hey we found a nice way to construct a large palette that does a pretty good job at covering all colors of some RGB space"
|
|
2023-03-03 09:20:44
|
the actual article is here: https://www.nature.com/articles/s41377-023-01073-x
|
|
2023-03-03 09:22:27
|
what surprises me is that this apparently got published in a journal that belongs to the Nature brand, which is supposedly quite reputable
|
|
2023-03-03 09:22:34
|
sigh
|
|
|
|
veluca
|
|
_wb_
what surprises me is that this apparently got published in a journal that belongs to the Nature brand, which is supposedly quite reputable
|
|
2023-03-03 09:23:50
|
I think it mostly has the reputation of being reputable, if that makes sense 😛
|
|
|
|
spider-mario
|
|
_wb_
Even that first sentence, what the... "Existing colour systems, such as RGB and CYMK, are all text-based" ?
|
|
2023-03-03 09:35:19
|
at first, I thought it might just be some weird formulation and tentatively gave them the benefit of the doubt, but then they fully doubled down:
> RGB (Red, Green, Blue), a light-colour system that contains 3×256 values of letter symbols
|
|
2023-03-03 09:35:55
|
----
> The letter-based colour codes make them difficult to explicitly express the relationship between colours and compute through mathematical operations, and ineffective in conversions among different colours.
what the hell is that supposed to mean
|
|
2023-03-03 09:40:15
|
> And because of the special property of prime numbers, each colour has only one code.
also true of (r, g, b) triplets due to the special property of 256²·r + 256·g + b
|
|
|
|
veluca
|
2023-03-03 09:41:53
|
did you already get to the mention of Goldbach's conjecture?
|
|
|
|
_wb_
|
2023-03-03 11:17:25
|
Getting a patent is even easier than getting a paper accepted in a high-ranking journal.
|
|
|
|
yoochan
|
2023-03-03 12:08:16
|
for a patent you "just" need to pay a patent office (and spend countless hours translating your technical innovation into an administrative gibberish)
|
|
|
|
Traneptora
|
2023-03-03 01:03:33
|
This strikes me as a theoretical thought experiment for mathemtical logic rather than a practical case
|
|
2023-03-03 01:04:02
|
but all they've done is say that the set of colors is countably infinite and provided a fairly obvious bijection to N
|
|
2023-03-03 01:05:10
|
I mean any math undergrad should be able to biject `N^k -> N` using a similar method
|
|
2023-03-03 01:05:48
|
it's like the cantor pairing function. theoretically interesting, practically uninteresting
|
|
|
|
_wb_
|
2023-03-03 01:51:28
|
they might even get a lot of citations when people reference their article to criticize what a load of bullsh*t it is
|
|
|
|
sklwmp
|
2023-03-03 02:18:24
|
This is annoying. My Acrobat Pro install can't handle indexed PNG with transparency; it just shows up as a black square if I have black text.
I have to specifically keep RGBA when optimizing the PNGs. Urgh.
|
|
|
HLBG007
|
2023-03-04 10:15:29
|
Do you know the problem, you tweet something and then your tweet liked first some bot profile that distributes his p*rn images? Terrible!
|
|
|
|
diskorduser
|
|
HLBG007
Do you know the problem, you tweet something and then your tweet liked first some bot profile that distributes his p*rn images? Terrible!
|
|
2023-03-04 01:25:51
|
Does it post p images in your tweet's replies?
|
|
|
|
HLBG007
|
|
diskorduser
Does it post p images in your tweet's replies?
|
|
2023-03-04 01:32:30
|
No, only like. The user profile self all content are flagged as nudity
|
|
|
|
spider-mario
|
|
|
Traneptora
|
2023-03-05 07:45:06
|
sending this to my brother who works in 3D arts
|
|
|
|
_wb_
|
2023-03-05 07:49:19
|
I wonder why there is such universal agreement between editors and viewers that images with alpha should be shown blended against a checkerboard background in light gray on gray.
|
|
2023-03-05 07:52:00
|
I meant, that's of course quite a good way to do it, it works for both bright and dark semitransparent stuff, but there are other things you could do. I wonder why and how they "standardized" on that particular pattern.
|
|
|
Fraetor
|
2023-03-05 09:11:39
|
It would be interesting to see where it originated. I imagine it was just copied from there afterwards.
|
|
|
|
DZgas Ж
|
|
spider-mario
|
|
2023-03-05 09:21:28
|
now fun
|
|
|
|
spider-mario
|
2023-03-05 09:30:25
|
I kind of liked the reflections/shadows in the original, but it’s probably not so easy to have (at least the reflections) with non-premultiplied alpha like in PNG
|
|
|
|
DZgas Ж
|
2023-03-05 09:55:31
|
it would be easier to recreate the whole effect from scratch with the original image. but it's easy to do
|
|
|
spider-mario
I kind of liked the reflections/shadows in the original, but it’s probably not so easy to have (at least the reflections) with non-premultiplied alpha like in PNG
|
|
2023-03-05 10:01:13
|
become invisible with black background.
|
|
2023-03-05 10:04:02
|
the color space is translated into alpha space 👆
|
|
|
jonnyawsom3
|
|
_wb_
I meant, that's of course quite a good way to do it, it works for both bright and dark semitransparent stuff, but there are other things you could do. I wonder why and how they "standardized" on that particular pattern.
|
|
2023-03-06 05:16:28
|
Some viewers tend to show pure black or pure white, the checkerboard is uncommon enough in images themselves so usually can't be mistaken for an image that has a solid background instead of transparent. Any replacement would need to be easy to see under most conditions but also rare in any normal use case
|
|
|
|
DZgas Ж
|
2023-03-07 07:58:40
|
I found 2 interesting projects for transmitting data by sound, these are **minimodem **and **quiet**. But I can not use them because they don't have ready builds (ha-ha-ha). But I found **AMODEM**, and it easy to use because - it is written in pure python
|
|
|
nec
|
2023-03-08 06:12:02
|
Don't you think modern video codecs are extremely impressive? I cut short video fragments to compare encoding and 10 frames quite often take only 100-200 kb, while a single png from that is already 1.5-8 mb.
|
|
|
|
jonnyawsom3
|
2023-03-08 06:42:02
|
That's because PNGs are lossless, the video codec is already throwing away detail and only records the differences in those 10 instead of having separate images
|
|
|
|
_wb_
|
|
nec
Don't you think modern video codecs are extremely impressive? I cut short video fragments to compare encoding and 10 frames quite often take only 100-200 kb, while a single png from that is already 1.5-8 mb.
|
|
2023-03-08 08:07:28
|
Inter frame can be like magic. Too bad it mostly works well at "web quality" kinds of fidelity. For "cinema quality", inter seems to be less useful...
|
|
|
|
jonnyawsom3
|
2023-03-08 08:19:28
|
The higher the framerate the better it does too, since 10 frames at 60fps would have 1/3 as much change as 20fps for example. It's why a lot of the time going from 30fps to 60 can only add on 1/10 the filesize
|
|
|
|
Traneptora
|
2023-03-08 08:21:49
|
assuming it was natively filmed at 60
|
|
2023-03-08 08:21:53
|
interpolation does not help
|
|
|
|
_wb_
|
2023-03-08 08:25:31
|
And assuming fidelity is low enough to have the motion estimation do most of the heavy lifting so there are few nonzero residuals left to encode
|
|
2023-03-08 08:28:23
|
The lower the fidelity, the more things are about the low-frequency signal and making the strong edges move smoothly, the better modern video codecs will work.
|
|
2023-03-08 08:31:41
|
But if it's about high fidelity, subtle textures and fine details, which are often poorly predicted by motion vectors, then I think modern video codecs are not really much better than intra-only.
|
|
|
|
Traneptora
|
2023-03-08 09:17:05
|
eh, motion vectors still work very well for low-frequency coefficients
|
|
2023-03-08 09:17:22
|
it's just the high-frequency coeffs that don't get helped as much
|
|
|
.vulcansphere
|
2023-03-09 01:04:04
|
Here's an AI-generated art of mine (Stable Diffusion 1.5, NightCafe Studio), with cyberpunk theme. Feel free to use for benchmarking, encoding test, etc (copyright-wise, I licensed it under CC-BY 4.0 so an attribution like VulcanSphere/CC-BY 4.0 should be enough (I assume AI art to be copyrightable in my country, Indonesia))
|
|
2023-03-09 01:04:19
|
|
|
|
|
jonnyawsom3
|
2023-03-09 08:43:53
|
Copyrighting AI art feels a bit redundant... Seeing as anyone could just type in a promt and get something similar
|
|
|
|
w
|
2023-03-09 10:16:35
|
same can be said about any kind of art
|
|
2023-03-09 10:16:42
|
anyone could just draw lines on a paper and get something similar
|
|
2023-03-09 10:19:26
|
it may be easier to compare it to literature - it's just words
|
|
2023-03-09 10:19:29
|
but in the end it's still art for reasons other than just the words
|
|
2023-03-09 10:23:23
|
i only just thought of that and it seems to work...
|
|
|
|
_wb_
|
2023-03-09 10:33:21
|
What complicates AI art is the question of to what extent it is derivative work (derived from the training set) and to what extent the training set images were properly licensed — I imagine that if the training set is only CC0 images, things are fine, but if there were thousands of CC-BY images in the training set, perhaps they should all be credited...
|
|
2023-03-09 10:39:40
|
But yes, whether something is art or not does not necessarily depend on how much 'artist effort' it required. Painting is much more labor-intensive than photography (which essentially can be just "pressing a button"), but both can be art — and both can be non-art as well, like when I get my walls painted 🙂
|
|
|
|
paperboyo
|
2023-03-09 10:44:30
|
> to what extent
Oh, that one’s easy. It’s 100.00% derivative. 😜
|
|
|
|
w
|
2023-03-09 10:46:26
|
but so is every kind of art
|
|
2023-03-09 10:46:49
|
everything is inspired by something
|
|
|
|
_wb_
|
2023-03-09 11:10:19
|
The question in my view is to what extent the concept "derivative" can get neutralized by dilution: if something is "derived" from 1 million other works at the same time, each with equal weight, is it still really a derived work?
I know there are examples of AI art where the actual original source artists can be quite recognizable, so then it's obviously "derivative". But I can also imagine there are cases where the contribution of each individual human source artist is a very small fraction of the total.
|
|
2023-03-09 11:12:16
|
There is a spectrum from "derived from" over "influenced by" to "part of the same overall culture but not really directly related".
|
|
2023-03-09 11:21:33
|
An artists who goes to a museum to view art by other artists, and then later makes art of their own, is not generally considered to be producing "derivative works".
Someone who makes a copy (whether faithful or reinterpreted) of an existing work is of course producing a derivative work. There is a spectrum between those two extremes and I think AI systems are somewhere on it but it's not clear where exactly — the answer may also depend on things like the specificity of the query or the architecture of the model.
|
|
|
|
jonnyawsom3
|
|
_wb_
The question in my view is to what extent the concept "derivative" can get neutralized by dilution: if something is "derived" from 1 million other works at the same time, each with equal weight, is it still really a derived work?
I know there are examples of AI art where the actual original source artists can be quite recognizable, so then it's obviously "derivative". But I can also imagine there are cases where the contribution of each individual human source artist is a very small fraction of the total.
|
|
2023-03-09 01:32:57
|
Many times when you can recognise a source image from the result, it's because the input was "By x" or "in the style of y".
If you tell a real artist to copy an image then they're not going to do much different :P
|
|
|
|
spider-mario
|
2023-03-09 02:22:25
|
allegedly, when the producer for Risky Business asked Tangerine Dream to make the soundtrack (https://youtu.be/V5aRgOk_dnU), they were asked to make it in the style of Steve Reich
|
|
2023-03-09 02:22:35
|
that pissed off Steve Reich quite a bit
|
|
2023-03-09 02:22:39
|
“They were paid to rip me off”
|
|
|
|
improver
|
|
w
everything is inspired by something
|
|
2023-03-09 11:37:09
|
that "something" is usually more like emotions & experiences, not literally copypasting others' stuff
|
|
2023-03-09 11:44:01
|
as someone who has so much trouble internalizing other's stuff when drawing, and instead ending up making new style changes every time i try to draw something by accidentally combining stuff by feel & just seeing how that goes, something as reductionist as that claim just hurts in all the wrong ways
|
|
2023-03-09 11:45:46
|
like you aint even doing art right if emotional expressions behind your technique, behind how your hand moves making up shapes aint yours
|
|
2023-03-09 11:49:03
|
i agree though that you can communicate your own message by describing layout of the scene in detail that symbolizes something out of your inner world
|
|
2023-03-09 11:49:37
|
like not all the things need to be experimental & expressionist
|
|
2023-03-09 11:51:35
|
anyways goodnight
|
|
|
|
w
|
2023-03-10 05:19:54
|
I think the different types of ai art should be looked at differently. The one where you draw something and it fills it in is one thing, but the popular stable diffusion where it takes a combination of words is another. I suggest it not be compared to as a painting, but instead literature. Because describing something is an art, and if you say otherwise, you do a disservice to writers.
|
|
|
|
.vulcansphere
|
2023-03-10 06:04:32
|
Interesting discussion about AI art here, I enjoyed reading differing opinions and views <:PepeOK:805388754545934396>
|
|
|
_wb_
An artists who goes to a museum to view art by other artists, and then later makes art of their own, is not generally considered to be producing "derivative works".
Someone who makes a copy (whether faithful or reinterpreted) of an existing work is of course producing a derivative work. There is a spectrum between those two extremes and I think AI systems are somewhere on it but it's not clear where exactly — the answer may also depend on things like the specificity of the query or the architecture of the model.
|
|
2023-03-10 06:05:36
|
Well said
|
|
2023-03-10 06:11:26
|
When ControlNet becomes more mainstream, the situation around AI art and copyrightability of it will be even more interesting (but still takes years to be resolved and even more in my country) 😂
|
|
|
|
diskorduser
|
2023-03-10 07:53:39
|
https://www.instagram.com/reel/Cl-9HXXsv-M/?igshid=YmMyMTA2M2Y=
|
|
|
|
afed
|
2023-03-10 01:30:26
|
some ai models even trying to generate signatures, because they were on training sources
https://www.artnews.com/wp-content/uploads/2022/12/Screen-Shot-2022-12-09-at-1.46.01-PM.png
|
|
|
|
jonnyawsom3
|
2023-03-10 01:43:30
|
Well yeah, it knows a signature is on most art, so it will try to make its own
|
|
|
|
spider-mario
|
2023-03-10 06:56:50
|
https://youtu.be/miomuSGoPzI
|
|
2023-03-10 06:57:17
|
context (?): https://youtu.be/Ppm5_AGtbTo
|
|
|
|
improver
|
2023-03-11 12:13:03
|
https://yogurt200.itch.io/heartlovepowertemple
|
|
|
|
afed
|
2023-03-12 11:03:50
|
https://www.reddit.com/r/Android/comments/11nzrb0/samsung_space_zoom_moon_shots_are_fake_and_here/
|
|
|
|
improver
|
2023-03-13 01:26:09
|
scamsung at it again
|
|
|
|
_wb_
|
2023-03-13 06:32:30
|
This is a logical continuation of the "AI enhancement" that phones are already doing for a while now
|
|
2023-03-13 06:35:18
|
In a way, this is relatively innocent. The moon is the moon — maybe if at some point it gets a new crater big enough to make a difference, the phone will need an update, but that's it. I am much more worried about AI enhancing photos taken on earth, where the bias in the training set can be way more problematic...
|
|
|
|
w
|
2023-03-13 07:17:58
|
the not innocent part was having influencers say brand B is fake and brand A is real, when brand A is just as fake as brand B
|
|
|
|
diskorduser
|
2023-03-13 07:19:23
|
But pixel phones use AI only for HDR processing, color correction and for recovering details right?
|
|
|
|
w
|
2023-03-13 07:19:52
|
i think majority of it is just denoising
|
|
2023-03-13 07:20:33
|
and all of those, but I dont know about "recovering details"
|
|
2023-03-13 07:24:29
|
(for default camera mode)
|
|
2023-03-13 07:24:46
|
im sure there's way more ai crap in portrait mode
|
|
|
|
improver
|
2023-03-13 07:45:42
|
i like my sony phon & i think they wouldn't pull stuff like that on me
|
|
|
|
_wb_
|
2023-03-13 08:08:35
|
|
|
2023-03-13 08:09:38
|
(figure is from a Sony patent: https://patents.google.com/patent/US8246454?oq=8246454)
|
|
|
|
veluca
|
|
diskorduser
But pixel phones use AI only for HDR processing, color correction and for recovering details right?
|
|
2023-03-13 08:09:46
|
As far as I understand AI is only used for features like magic eraser... And probably zooms beyond the optical zoom, but I don't know for sure and if I did know for sure I very likely shouldn't tell 🤣
|
|
|
veluca
As far as I understand AI is only used for features like magic eraser... And probably zooms beyond the optical zoom, but I don't know for sure and if I did know for sure I very likely shouldn't tell 🤣
|
|
2023-03-13 08:11:01
|
Although possibly not, 30x zoom can and will allow me to read text and AI superresolution is absolutely terrible at that
|
|
|
|
improver
|
|
_wb_
(figure is from a Sony patent: https://patents.google.com/patent/US8246454?oq=8246454)
|
|
2023-03-13 08:13:55
|
both evil & funny at once. i suspect its their gaming console or tv division so not exactly the same
|
|
2023-03-13 08:14:30
|
kinda like i don't consider samsung's ssds to be rotten as much as phones or tvs
|
|
|
|
_wb_
|
2023-03-13 08:15:43
|
yeah, I suppose different divisions of these huge conglomerates can be quite different in terms of quality and ethics, even when they're sharing the same brand name
|
|
2023-03-13 08:16:05
|
like not all Google departments are the same 🙂
|
|
|
|
DZgas Ж
|
|
_wb_
|
|
2023-03-13 11:42:58
|
holy
|
|
|
|
Fraetor
|
|
veluca
As far as I understand AI is only used for features like magic eraser... And probably zooms beyond the optical zoom, but I don't know for sure and if I did know for sure I very likely shouldn't tell 🤣
|
|
2023-03-13 07:46:21
|
That optical zoom was probably "AI" at one point, but when AI becomes especially useful it stops being called AI.
|
|
|
|
jonnyawsom3
|
2023-03-13 07:52:55
|
My phone was meant to be the first with a 2x 'Lossless zoom', in reality I notice the 20MP option doesn't allow any zoom, and the jpeg has an extra thousand pixels over the RAW seeming out of thin air
|
|
|
gb82
|
|
veluca
As far as I understand AI is only used for features like magic eraser... And probably zooms beyond the optical zoom, but I don't know for sure and if I did know for sure I very likely shouldn't tell 🤣
|
|
2023-03-13 11:31:21
|
I believe the iPhone uses AI for zoom beyond optical. The pixel uses Super Res Zoom which is like sensor shift on high end cameras to interpolate detail
|
|
|
My phone was meant to be the first with a 2x 'Lossless zoom', in reality I notice the 20MP option doesn't allow any zoom, and the jpeg has an extra thousand pixels over the RAW seeming out of thin air
|
|
2023-03-13 11:32:08
|
if your phone has a >48mp sensor (like the Pixel 7) it's simply doing a sensor crop & not doing pixel binning most likely. not much trickery going on there
|
|
2023-03-13 11:32:13
|
the raw probably bins no matter what
|
|
2023-03-13 11:32:21
|
that's how it works on my Pixel anyhow
|
|
|
|
veluca
|
2023-03-13 11:36:09
|
I probably could ask relevant people for more details 😛 but I guess I'm not too curious
|
|
|
|
jonnyawsom3
|
2023-03-13 11:36:21
|
My phone came out a month after the Pixel 2, so I doubt that :P
|
|
2023-03-13 11:37:12
|
I should've clarified, 20MP is the maximum setting, not just 'an option'
|
|
2023-03-13 11:38:17
|
November 2017
|
|
2023-03-13 11:40:24
|
It was also supposedly one of the first phones to have AI hardware acceleration with camera recognition and all that. Good ol' Huawei Mate 10 Pro
|
|
|
|
gb82
|
|
It was also supposedly one of the first phones to have AI hardware acceleration with camera recognition and all that. Good ol' Huawei Mate 10 Pro
|
|
2023-03-13 11:43:23
|
oh gotcha
|
|
2023-03-13 11:43:37
|
yeah def doesn't do the sensor crop binning
|
|
|
|
jonnyawsom3
|
2023-03-13 11:45:27
|
Weirdly it got a firmware update a few months ago after 3 years of nothing. Completely changed the UI and the camera app removing half the options so I can't even pick a resolution anymore and video image stabilisation is permanently on. Very weird
|
|
|
|
gb82
|
2023-03-13 11:46:12
|
that's wack
|
|
|
|
afed
|
2023-03-14 08:01:37
|
https://techcrunch.com/2023/03/14/openai-releases-gpt-4-ai-that-it-claims-is-state-of-the-art/
|
|
|
|
spider-mario
|
2023-03-14 08:46:02
|
|
|
2023-03-14 08:48:37
|
(original track: https://youtu.be/_EDV5yBzk3g)
|
|
|
|
_wb_
|
2023-03-14 08:54:48
|
Is that you? Did you cut your hair?
|
|
|
|
spider-mario
|
2023-03-14 09:09:58
|
it is, and I did 😄
|
|
|
|
diskorduser
|
2023-03-16 06:33:50
|
https://twitter.com/emollick/status/1635700173946105856?s=20
|
|
|
|
sklwmp
|
2023-03-18 11:49:05
|
https://twitter.com/ItsSimonTime/status/1636857478263750656
|
|
|
retr0id
|
2023-03-18 11:51:41
|
it me
|
|
2023-03-18 11:51:52
|
I'm in the picture
|
|
|
|
jonnyawsom3
|
2023-03-18 04:24:10
|
1. How on earth did that happen
2. Just how big is that 'cropped' image file compared to original/restored?
|
|
|
|
sklwmp
|
2023-03-18 04:31:43
|
from what i can gather from the scattered posts, basically the cropped image is written over the original, but something changed in the API that made it so that the file isnt truncated
meaning that some of the image data is still kept even after overwriting with the cropped version
|
|
|
|
zamfofex
|
2023-03-18 11:00:36
|
This seems relevant: <https://twitter.com/DontStealMacOSY/status/1636947728096956417>
|
|
|
username
|
|
sklwmp
https://twitter.com/ItsSimonTime/status/1636857478263750656
|
|
2023-03-18 11:02:42
|
https://www.da.vidbuchanan.co.uk/blog/exploiting-acropalypse.html
|
|
|
|
Traneptora
|
2023-03-19 12:20:00
|
why would it not be truncated by default if opened for writing?
|
|
2023-03-19 12:20:04
|
that... doesn't make sense
|
|
2023-03-19 12:20:15
|
as a developer I would assume that it would truncate the file
|
|
|
|
sklwmp
|
2023-03-19 12:21:24
|
i don't know either... it just seems stupid
|
|
|
|
zamfofex
|
2023-03-19 03:16:07
|
I talked about text shaping to a friend of mine, and he seemed confused about it, so I wrote a bit about various text concepts and how they are represented in computers to explain it to him. I wanted to avoid using too many technical terms, so I wrote about it in a very informal way. It seems instructive enough, though, so I felt like I should share it here, in case anyone feels like it might be interesting to read: <https://gist.github.com/zamfofex/948b0a51a0ed6ee4b90d82f41585d1d3> (I might move it to my blog eventually, though.)
|
|
|
|
diskorduser
|
2023-03-19 02:58:20
|
https://twitter.com/AstraKernel/status/1637346219252154369?t=GIcmwBXE2Nia5mCMew0Nmw&s=19
|
|
|
|
jonnyawsom3
|
2023-03-19 08:30:05
|
Dug out the old family Windows XP machine that got put in a closet because it ran out of disk space and was too slow.
Turns out it was a single 50GB video file using up 2/3 of the drive, and 98% fragmentation too.
Gonna love seeing how well it compresses into x264 :P
|
|
2023-03-19 08:30:58
|
Also got a few bitmaps and jpegs to test JXL with too
|
|
|
Dug out the old family Windows XP machine that got put in a closet because it ran out of disk space and was too slow.
Turns out it was a single 50GB video file using up 2/3 of the drive, and 98% fragmentation too.
Gonna love seeing how well it compresses into x264 :P
|
|
2023-03-20 12:00:16
|
Turns out it was horrendously corrupted along with 40 other videos, which makes sense considering it was rare to get a 64MB flash card yet alone 64GB back then
|
|
|
|
improver
|
2023-03-20 12:01:17
|
rip.
|
|
2023-03-20 12:02:01
|
probably demagnetized
|
|
|
|
jonnyawsom3
|
2023-03-20 12:04:35
|
Well, either they're all corrupted or they just so happen to be around 3 seconds long before throwing infinite `ac-tex damaged` errors. A few are still a minute or so long
|
|
2023-03-20 12:05:30
|
Weird thing is one of the corrupted files plays half a frame of the next file in the folder, makes me wonder if the old camera did some jank of combining videos or splitting them up
|
|
|
|
improver
|
2023-03-20 12:06:08
|
hard drives do lose their magnetic charge with time & that next frame behavior sounds like FS metadata corruption too
|
|
|
|
jonnyawsom3
|
2023-03-20 12:09:24
|
Actually... It looks like there's some kind of marker between the 'actual' video and the next file that cuts into it... Interesting
|
|
|
|
improver
|
2023-03-20 12:10:40
|
the dead drive explorer
|
|
|
|
jonnyawsom3
|
2023-03-20 12:11:22
|
Thinking about it, I could've just posted the errors, you're almost certainly right
`Packet corrupt (stream = 1, dts = 301500).
ac-tex damaged at 22 6
Warning MVs not available
concealing 987 DC, 987 AC, 987 MV errors in I frame`
|
|
2023-03-20 12:25:49
|
Oh, actually, now I remember seeing an end of file error too embedded inside the metadata info for some reason
|
|
|
|
diskorduser
|
2023-03-20 01:00:19
|
https://www.ghacks.net/2023/03/19/yandex-browsers-gpu-crashes-dropped-significantly-after-pretending-to-be-chrome/
|
|
|
|
jonnyawsom3
|
|
|
DZgas Ж
|
2023-03-20 07:52:14
|
My graph of the dependence of the image size and the amount of empty space (in characters) due to line breaks (end of paragraph) and word breaks when there is not enough space (affects small resolutions) - on standard text when using a monospaced font with a width of 8 pixels
|
|
2023-03-20 08:03:46
|
I didn't find any software except windows paint and GIMP that has a textbox For word wrapping text when there is a lack of resolution. literally not found. GIMP adds one unnecessary pixel between the lines. unfortunately, windows paint is not suitable because it does text smoothing and destroys pixel fonts. bullshit <:SadOrange:806131742636507177>
|
|
|
fab
|
2023-03-21 10:03:57
|
https://github.com/virtualvinodh/satisarsharada
|
|
2023-03-21 10:58:19
|
https://www.cufonfonts.com/font/rileno-sans
|
|
2023-03-21 11:00:20
|
https://github.com/ThePitagon/pitagon-sans
|
|
|
|
zamfofex
|
|
fab
https://github.com/ThePitagon/pitagon-sans
|
|
2023-03-21 05:05:35
|
This one reminds me a loto of the Ubuntu fonts in their primary approach/ideas!
|
|
|
|
fab
|
2023-03-21 05:54:03
|
Yes
|
|
2023-03-21 05:54:18
|
Try the variable ttf
|
|
2023-03-21 05:54:24
|
At 10point
|
|
2023-03-21 05:54:28
|
On notepad
|
|
|
|
diskorduser
|
2023-03-22 02:37:48
|
https://www.gizchina.com/2023/03/20/amazon-prepares-a-google-chrome-killer-to-bring-justice-to-your-ram/
|
|
|
|
w
|
2023-03-22 02:48:36
|
is this post ai generated
|
|
|
|
gb82
|
2023-03-22 04:10:42
|
gptzero.me
|
|
2023-03-22 04:10:46
|
give it a shot
|
|
|
|
_wb_
|
2023-03-22 06:27:31
|
https://twitter.com/juanbuis/status/1638292590842740737?t=UyQ5IZXcNOh-5w5Gv-uDrQ&s=19
|
|
|
|
jonnyawsom3
|
2023-03-22 11:24:27
|
Only when resaving a preexisting file, screenshots aren't affected
|
|
|
|
gb82
|
2023-03-22 07:16:41
|
google's bot like jxl
|
|
|
|
_wb_
|
|
|
diskorduser
|
2023-03-24 02:46:51
|
https://www.youtube.com/@LinusTechTips/videos
|
|
2023-03-24 02:47:26
|
his account got hacked.
|
|
2023-03-24 05:14:04
|
Looks like they fixed it now.
|
|
|
|
Tirr
|
2023-03-24 06:22:23
|
https://github.blog/2023-03-23-we-updated-our-rsa-ssh-host-key/
|
|
|
|
diskorduser
|
2023-03-24 10:23:28
|
https://www.youtube.com/watch?v=yGXaAWbzl5A
|
|
|
|
sklwmp
|
2023-03-24 01:45:21
|
TL;DW: session hijacking
one of their non-video prod employees opened a malicious email attachment, malware got access to browser data, session cookie hijacked
|
|
2023-03-24 01:45:30
|
apparently it's a common hacking thing on youtube right now
|
|
|
|
diskorduser
|
2023-03-24 02:59:57
|
https://twitter.com/sigma_da_enigma/status/1533094947158900736?s=20
|
|
|
|
gb82
|
2023-03-25 12:06:02
|
It seems like Windows was the point of failure after all
|
|
|
|
_wb_
|
2023-03-25 08:23:35
|
https://www.intel.com/content/www/us/en/newsroom/news/gordon-moore-obituary.html?cid=iosm&source=twitter&campid=hq_ao_q1-2023&content=100003943892992&icid=reactive-engagement&linkId=100000196298440
|
|
|
|
yoochan
|
2023-03-25 08:54:29
|
Requiescat in pace
|
|
|
|
_wb_
|
2023-03-25 09:07:42
|
Someone on twitter said they'll move him every two years to a coffin that is twice as small...
|
|
|
|
yoochan
|
2023-03-25 09:32:45
|
😆
|
|
2023-03-25 05:46:03
|
reading an article about exotic distros (linux) I found an interesting one where each library is installed in a special folder which bears the version number, which allows to have many revision of the same lib in parallel (patchelf is used to rewrite the RPATH to make everything works if I remember...) yet, I'm unable to find its name again, nor the article... I only remember it was not a POSIX distribution... if it rings a bell for someone...
|
|
|
|
190n
|
2023-03-25 05:58:29
|
gobo linux?
|
|
|
|
yoochan
|
2023-03-25 05:58:47
|
yes !!
|
|
2023-03-25 05:59:12
|
thank you very much 🙂
|
|
|
|
zamfofex
|
2023-03-25 09:08:19
|
There is also Guix, which I think has more packages! <https://guix.gnu.org>
|
|
|
|
yoochan
|
|
There is also Guix, which I think has more packages! <https://guix.gnu.org>
|
|
2023-03-26 06:35:30
|
I'll have a look. I'm a bit disappointed by gobo, it seems to deal with version selection via symbolic links instead of rpath fiddling as I remembered... Which is meh 🤔
|
|
|
|
Traneptora
|
2023-03-26 07:36:02
|
nixos is another one
|
|
|
|
spider-mario
|
2023-03-26 07:56:47
|
I am surprised that gobo linux was the correct answer – nixos is indeed the first that came to my mind
|
|
2023-03-26 07:56:57
|
I hadn't even heard of gobo
|
|
2023-03-26 07:57:30
|
whereas I have 4 friends using nix (but maybe there's a network effect here)
|
|
|
|
yoochan
|
2023-03-26 09:51:31
|
I'll have a look at nixos too
|
|
2023-03-26 09:55:06
|
but it's nice to see that by searching a single distro I found 3 😄
|
|
|
|
_wb_
|
2023-03-26 04:09:40
|
https://twitter.com/erlichya/status/1639973591214182400?t=P1qYBGH9fs7eq5xRcJAZHQ&s=19
|
|
2023-03-26 04:10:02
|
L and Z both died
|
|
|
|
gb82
|
2023-03-26 06:35:38
|
<:FeelsSadMan:808221433243107338>
|
|
|
|
190n
|
2023-03-26 07:08:10
|
<:PepeSad:815718285877444619>
|
|
|
|
yoochan
|
2023-03-26 07:31:01
|
no joke about the size of their coffins ?
|
|
|
|
improver
|
2023-03-26 08:59:59
|
<:PepeSad:815718285877444619>
|
|
|
|
Jyrki Alakuijala
|
|
yoochan
no joke about the size of their coffins ?
|
|
2023-03-27 02:52:26
|
Z's obituary: see L's obituary
|
|
|
|
yoochan
|
2023-03-27 06:05:17
|
Approved 😁
|
|
|
|
_wb_
|
|
Jyrki Alakuijala
Z's obituary: see L's obituary
|
|
2023-03-27 06:07:54
|
Z's obituary: see L`<distance: 18, length 11>`
|
|
|
|
DZgas Ж
|
2023-03-27 09:18:09
|
damn, it was very interesting to combine 1 million images into 18 thousand long manga images, process them with a neural network for remove dots, then reduce the size to the size of a jpeg xl macroblock 256px (768px) and then compress it with a special cjxl build with multi-layer progressive decoding
|
|
2023-03-27 09:59:06
|
there is a laughable fuck everywhere, they don't know how to compile a project for all processor commands together and do it separately for SSE and for AVX builds and that would, ahahah, that would not be SHAME - write "AVX-only". THEY CAME up with writing 64bit-v3 ahahah
|
|
|
|
improver
|
2023-03-27 10:31:44
|
you sure like your old hardware don't you
|
|
|
|
Nova Aurora
|
|
yoochan
no joke about the size of their coffins ?
|
|
2023-03-28 01:52:09
|
They are getting DEFLATED
|
|
|
|
DZgas Ж
|
|
improver
you sure like your old hardware don't you
|
|
2023-03-28 07:04:34
|
"old" for what?
|
|
2023-03-28 07:05:21
|
My hardware is working, and it is not written anywhere that it is outdated.
|
|
2023-03-28 07:10:44
|
64bit is legacy compatibility. you just build the software for all processor instructions. all modern programs work like this. Something I don't see in libjxl - different build for sse3-only for avx-only for win xp for win 11 -- no, it just works everywhere.
|
|
|
|
zamfofex
|
2023-03-28 07:49:03
|
Imma agree with you there, at least partially. I really dislike when people claim my computer specs are “old” whereas my computer has been serving me completely fine for several years now. Though I understand technology moves on and people want to be able to use the latest hardware improvements actively. I think it’s fine, as long as they provide more widely compatible binaries too.
|
|
2023-03-28 07:50:42
|
Generally, newer CPUs provide more instructions that can help with performance. Although it’s not an essential thing, certain programs benefit from it quite a lot.
|
|
|
|
improver
|
2023-03-28 08:08:45
|
age is independent of how good it is, working well does not make it not old relative to the most of current machines
|
|
2023-03-28 08:09:27
|
as much as you want to twist it, words have meanings and old isn't the same as 'obsolete'
|
|
2023-03-28 08:10:39
|
it wasn't an attack, just an observation
|
|
|
|
zamfofex
|
2023-03-28 08:13:03
|
That’s fine, though some people do claim that old things are obsolete. In the sense that some people will say something is “old” to mean it shouldn’t be used anymore. At any rate, age is a fairly relative concept. Depending on the context, 5 years might be a long time, but sometimes it might also be very little. Sometimes it just depends on how people choose to perceive it, and general perception might change over time.
|
|
|
|
improver
|
2023-03-28 08:21:51
|
arguably in case of computers being old also equates to being inefficient power usage wise. but, like, obsolescence is a concept of what owner of the device decides, if they don't care about that then it's fine.
|
|
2023-03-28 08:26:49
|
more interesting thing is what one considers 'old' is shaped relative to what one's peers have
|
|
2023-03-28 08:28:43
|
though in the age of the internet this sorta have mostly lost old ways of how it worked
|
|
2023-03-28 10:02:07
|
|
|
|
gameplayer55055
|
2023-03-28 10:20:40
|
mouse is better than keyboard
|
|
|
|
diskorduser
|
2023-03-28 10:44:57
|
win10 enterprise - best OS.
|
|
2023-03-28 10:45:56
|
Yeah that thing
|
|
|
|
gameplayer55055
|
|
diskorduser
win10 enterprise - best OS.
|
|
2023-03-28 11:03:37
|
the best os is Emacs <:trol:1005506850369831033>
|
|
2023-03-28 11:05:08
|
but actually there's no thing like best windows version. in xp you had problems with drivers, in 7 you had codecs problems, in 10 you have different apps that you didn't ask for
|
|
2023-03-28 11:05:36
|
linux? everyone may make the OS best for himself
|
|
|
|
improver
|
2023-03-28 11:06:59
|
~~Windows XP 64-bit edition~~
|
|
|
|
gameplayer55055
|
|
improver
~~Windows XP 64-bit edition~~
|
|
2023-03-28 11:09:05
|
windows pos, on your ATMs, for years, forever
|
|
|
|
afed
|
2023-03-28 01:28:08
|
also, many apps now depend on chromium (electron/cef) and if chromium does not support some hardware or oses, then most likely these apps will not have support either
https://help.steampowered.com/en/faqs/view/4784-4F2B-1321-800A
|
|
2023-03-28 01:33:40
|
<https://www.electronjs.org/blog/electron-23-0>
|
|
|
|
diskorduser
|
|
gameplayer55055
but actually there's no thing like best windows version. in xp you had problems with drivers, in 7 you had codecs problems, in 10 you have different apps that you didn't ask for
|
|
2023-03-28 01:45:55
|
Windows 10 enterprise ltsc doesn't have a lot of apps.
|
|
|
|
yoochan
|
|
DZgas Ж
damn, it was very interesting to combine 1 million images into 18 thousand long manga images, process them with a neural network for remove dots, then reduce the size to the size of a jpeg xl macroblock 256px (768px) and then compress it with a special cjxl build with multi-layer progressive decoding
|
|
2023-03-29 07:00:33
|
I'm eager to see the result and interested in understanding a bit better your goal here ?...
|
|
|
|
DZgas Ж
|
|
yoochan
I'm eager to see the result and interested in understanding a bit better your goal here ?...
|
|
2023-03-29 08:04:05
|
Hah. Well. When it's ready. You will be able to see it on hentai.i2p
|
|
|
|
yoochan
|
2023-03-29 08:07:14
|
lol
|
|
2023-03-29 08:07:35
|
but what are you *trying* to do ? put a whole manga on a single file ?
|
|
|
|
DZgas Ж
|
|
yoochan
but what are you *trying* to do ? put a whole manga on a single file ?
|
|
2023-03-29 08:56:36
|
I've already done it
|
|
|
|
yoochan
|
2023-03-29 08:59:00
|
one page per frame ?
|
|
2023-03-29 08:59:14
|
each one on top of each other ?
|
|
|
|
DZgas Ж
|
2023-03-29 08:59:44
|
|
|
2023-03-29 09:01:11
|
|
|
2023-03-29 09:01:44
|
It takes 1 gigabyte of ram
|
|
|
|
yoochan
|
2023-03-29 09:06:00
|
lol, I expected one page per frame in fact 😄 how do you deal with dual page wide drawings ?
|
|
2023-03-29 09:06:21
|
the decoder could do cropped decoding on the fly...
|
|
2023-03-29 09:11:06
|
but your format would be very convenient for this new kind of graphical novel optimized for smartphone which are read as a continuous vertical strip
|
|
2023-03-29 09:11:23
|
(I think it was korean stuff initially)
|
|
|
|
gameplayer55055
|
|
DZgas Ж
|
|
2023-03-29 09:19:41
|
gonna send to my teacher
|
|
2023-03-29 09:20:00
|
university disk is already full of 28mp images from dslr
|
|
|
|
gb82
|
2023-03-31 06:59:54
|
https://www.lensrentals.com/blog/2023/03/how-your-camera-and-image-processor-determine-colors/
|
|
2023-03-31 07:03:19
|
wait was this already sent
|
|
|
|
yoochan
|
2023-03-31 07:14:35
|
<@456226577798135808> does this nickname comes from bofuri ? a lovely anime IMO
|
|
|
|
diskorduser
|
2023-03-31 07:29:37
|
bofuri nice anime
|
|
|
|
gameplayer55055
|
2023-04-01 09:27:42
|
i heard Microsoft Windows is about to add Wayland support to boost gaming performance
|
|
|
|
improver
|
2023-04-01 10:31:09
|
wewlad support
|
|
2023-04-01 10:31:31
|
|
|
2023-04-01 10:31:37
|
one human not included
|
|
|
|
diskorduser
|
2023-04-02 12:45:07
|
https://open.spotify.com/track/2IFiZRrjyj9vVos39glK9j?si=0b19f761b3e24d81
|
|
|
|
improver
|
2023-04-02 02:25:28
|
https://spotify.link/6EA6xARIFyb
|
|
2023-04-02 11:22:31
|
https://youtu.be/hnIdFDzpo_o
|
|
|
|
fab
|
2023-04-03 12:14:54
|
https://www.fontspace.com/noodle-grotesque-font-f94571
|
|
|
|
diskorduser
|
2023-04-03 12:54:25
|
Fab grotesque when?
|
|
|
|
gameplayer55055
|
2023-04-03 02:30:31
|
|
|
2023-04-03 02:31:26
|
honestly: I don't like when laptop becomes a spider 🕷️, wires must be a thing of PCs
but hopefully magsafe hdmi jack and sdcard is back
|
|
2023-04-03 07:28:50
|
https://nullprogram.com/blog/2021/12/30/
|
|
|
|
improver
|
2023-04-03 09:41:36
|
oh yes unicode stuff of windows is a big lole
|
|
|
|
gb82
|
2023-04-03 09:57:13
|
|
|
2023-04-03 10:18:30
|
|
|
|
|
DZgas Ж
|
|
DZgas Ж
damn, it was very interesting to combine 1 million images into 18 thousand long manga images, process them with a neural network for remove dots, then reduce the size to the size of a jpeg xl macroblock 256px (768px) and then compress it with a special cjxl build with multi-layer progressive decoding
|
|
2023-04-05 12:50:37
|
It turned out that OpenCV has big problems with opening images like 200,000 pixels. Very bad
|
|
|
|
gameplayer55055
|
2023-04-05 01:12:57
|
who the f uses such images. tho would be useful in 3d graphics for texture stitching, meteorology and telescopy?
|
|
|
|
_wb_
|
2023-04-05 04:59:05
|
I was here today. Nice view from over there...
|
|
|
|
DZgas Ж
|
|
gameplayer55055
who the f uses such images. tho would be useful in 3d graphics for texture stitching, meteorology and telescopy?
|
|
2023-04-05 11:46:46
|
It sounds like excuses instead of solving problems
|
|
2023-04-05 11:51:37
|
because I can't work with such images, not because "you don't have enough memory" and not because "it's very slow" - but because software programmers didn't give a fuck
to make good code
and of course it surprises me in such projects that are used by millions of people around the world.
|
|
2023-04-05 11:55:53
|
Here's a good example, CJXL, because I just asked my friend (which has a lot of RAM) to encode an image of 768x130000, which equals 100 megapixels, and he was just able to do it. And I didn't have any problems decoding it, yes, it took 2 gigabytes of memory just to view it, and about a minute to decode it, but it JUST works ☝️ <:JXL:805850130203934781>
|
|
|
Cool Doggo
|
|
DZgas Ж
because I can't work with such images, not because "you don't have enough memory" and not because "it's very slow" - but because software programmers didn't give a fuck
to make good code
and of course it surprises me in such projects that are used by millions of people around the world.
|
|
2023-04-05 11:58:42
|
millions of people use it because they aren't trying to view the largest images possible
|
|
|
|
DZgas Ж
|
|
Cool Doggo
millions of people use it because they aren't trying to view the largest images possible
|
|
2023-04-05 11:59:29
|
when a program does not perform its tasks, people use another program.
|
|
2023-04-06 12:00:11
|
in my case, I had to write a pillow "streaming" in png (on python)
|
|
2023-04-06 12:00:36
|
line by line
|
|
|
|
Cool Doggo
|
2023-04-06 12:01:10
|
you do understand what you are trying to do is not what everyone else is trying to do right
|
|
|
|
DZgas Ж
|
|
Cool Doggo
you do understand what you are trying to do is not what everyone else is trying to do right
|
|
2023-04-06 12:05:19
|
I don't consider this an argument. An argument for me would be an inscription on the program's website that says "unfortunately our software CANNOT work with resolutions of 200,000 pixels sorry"
And what you say has no value - programs are made to solve problems, the fact that My problem is different from other people's problems is not my care.
The program that should work with images, But it does not cope with this task, and what conclusion can be drawn - this is shit
|
|
|
|
Traneptora
|
2023-04-06 12:05:54
|
why would it say that?
|
|
2023-04-06 12:06:27
|
I agree that it would be a nice feature for libjxl to have, as one of the nice things that JPEG XL can do is region-of-interest decoding for massive massive images
|
|
|
|
DZgas Ж
|
2023-04-06 12:06:52
|
it's about OpenCV
|
|
|
|
Traneptora
|
2023-04-06 12:06:54
|
but it doesn't need a disclaimer that says the thing that nobody expects to work doesn't in fact work
|
|
2023-04-06 12:07:23
|
OpenCV is for computervision, right? there's literally no reason to use that on massive images
|
|
|
|
DZgas Ж
|
2023-04-06 12:07:38
|
I don't have the slightest complaint about JPEG XL
|
|
|
Traneptora
OpenCV is for computervision, right? there's literally no reason to use that on massive images
|
|
2023-04-06 12:08:15
|
well again say the same thing as the person a couple of messages ago
|
|
|
|
Traneptora
|
2023-04-06 12:09:13
|
probably because what you are trying to is an objectively bad idea
|
|
|
|
DZgas Ж
|
2023-04-06 12:09:18
|
<@853026420792360980>Let's do this, you show me where on the project website, or in the documentation it is written that this is the maximum image size with which the program can work - then we'll talk
|
|
|
|
Traneptora
|
2023-04-06 12:09:27
|
there is no maximum image size
|
|
2023-04-06 12:09:29
|
it's limited by your RAM
|
|
2023-04-06 12:09:34
|
you just happen to have low-end hardware
|
|
|
|
DZgas Ж
|