Hacker News Reader

Comment by flashback2199 | Hacker News Reader

> 2D image library

Maybe I'm ignorant but what's a 1D image ?

I'd highly recommend reading Flatland. Not only because you had to ask this, but because it's a fun (and short) read.

LOL yes I've read Flatland. I was trying to sus out why the parent was specifying 2D image since that's the default interpretation of "image"...

> > 2D image library > Maybe I'm ignorant but what's a 1D image ?

There is no such thing ("2D image" is still useful, to distinguish from 3D.)

> "2D image" is still useful, to distinguish from 3D.

No it isn't? You don't say "here's that 2D image you wanted" when you send someone a jpg.

still useful in contexts like the one it was used in upthread, which was not "when you send someone a jpeg".

There is absolute such as thing as a 1d image, in many disciplines.

(+ the meaning in math as what comes out of a function....)

There are even 1D cameras - e.g. scanners, finish line cameras for races, some old barcode scanners (although most can do 2D now and even those that only support 1D barcodes might make use of a 2D camera).

Without the "2D" qualifier, it might refer to something like a disk image.

A line of pixels.

Ah ok. It sounds like it's important to distinguish the two by saying 2D image, but I'm sure the reason is too advanced for me.

More likely to be distinguishing 2D from 3D graphics, and just saying "image" can be ambiguous when firmware images are in play.

So, are video-files of 2D images technically 3D then?

and time-based volumetric recording of 3D video-games are 4D?

Time doesn't necessarily have to be another dimension. They're just what they are but organized in time.

You don't have to think of it as a dimension if that obfuscates the problem or isn't a useful mental model for you, but "organizing" it is what a dimension does. You can think of an image as colors or intensities "organized by" (I would call it "indexed by") width and height. If you work with videos in machine learning, you generally accept a 6 dimensional tensor (width, height, time, red, green, blue - your order may vary, time often comes first. And you may use grayscale instead of color to reduce the number of dimensions.).

> If you work with videos in machine learning, you generally accept a 6 dimensional tensor (width, height, time, red, green, blue

(Assuming you do work with ML+videos) - it's surprising to hear you say you work with RGB instead of YUV - can you briefly explain how that's the case? I'd have thought that using luma/chroma separation would be much easier to work with (not just with traditional video tooling, but ML/NNs/etc themselves would have an easier time consuming it.

To clarify, I don't work professionally with videos, I've hacked on some projects and read some books about it. My professional experience with ML models is in writing backends to integrate with them, the models I've designed/trained were for my own education (so far, at least). The answer to your question is probably, "I'm a dilettante who doesn't know better, you may well know more than me."

I take the impression that much of the time, color doesn't provide much signal and gives your model things to overfit on, so you collapse it down to grayscale. (Which is to say, most of the time you care about shape, but you don't care about color.) But I bet there are problem spaces where your intuition holds, I'm sure that there's performance the be wrung out of a model by experimenting with different color spaces who's geometry might separate samples nicely.

I did something similarish a few months ago where I used LDA[1] to create a boutique grayscale model where the intensity was correlated to the classification problem at hand, rather than the luminosity of the subject. It worked better than I'd have guessed, just on it's own (though I suspect it wouldn't work very well for most problems). But the idea was to preprocess the frames of the video this way and then feed it into a CNN [2]. (Why not a transformer? Because I was still wrapping my mind around simpler architectures.)

[1] https://en.wikipedia.org/wiki/Linear_discriminant_analysis

[2] https://en.wikipedia.org/wiki/Convolutional_neural_network

> So, are video-files of 2D images technically 3D then?

Video files are a linked list of 2D images.

> and time-based volumetric recording of 3D video-games are 4D?

Kind of, but it would be an oversimplification. Typically when we refer to the dimensionality of objects, we're referring to physical dimensions. Time is a temporal dimension. I think it would be more specific to say this is a linked list of 3-dimensional images, right?

You can render a video file as a volume. I've looked at using that to make a video compression algorithm that operated on the volume rather than on the 2D frame stream. My hunch was that shapes in the 3D volume changed more predictably than surfaces from frame to frame because the frames describe the movement of objects in space. But they're projected onto a two dimensional surface. So you get these interesting 3D shapes that have fairly predictable qualities across larger spans of time than your average 2D encoder sees while encoding a video. But I never could get it to work more efficiently than existing algorithms.

Still, it was a fun project.

existing algorithms do a lot of compression across frame-sequences, but yeah not quite in the same way as the imputed volume.

I wonder if your idea would work for lightfield captures, or time sequences of a lightfield.

I suspect you'd need your compression to "understand" the object relationships and camera movement to do better than frame sequences, and it'd probably still be incredibly hard because you then add a lot of extra information first in the hope they let you discard more pixel data...

But the more you understand the scene, the more you can potentially outright reconstruct, and in some contexts more loss would be entirely fine if the artifacts are plausible.

That's exactly where I ended with this: I was decomposing the scene and realized that if you had the ability to do that reliably enough you'd be recreating a model rather than an image and then re-rendering that model. But at that point I don't think you are looking at a compression algorithm any more other than in the very broadest sense of the word. Boundaries between objects would start to look fuzzy otherwise. As in: you'd no longer know exactly where the table ended and the hand started unless you modeled it precisely enough and at that point you have an object model. So you might as well use it to render the whole scene.

Note that I did this in '98 or so, when there was less of a computational budget, maybe what I couldn't hack back then is feasible today.

2D images are already 3D if you consider the colors (rgba) a dimension or the image contains layers like gif

Pixels are a unit of area like an acre or square meter (see [1] if skeptical). So a line of pixels is still 2D, in the same way a 1x5 unit rectangle is a 2D object with an area of 5 units squared. I'm not sure there's an accepted name for a 1 dimensional picture element. Maybe lenxel, working backwards from length like voxels works backwards from volume?

I like the sibling's suggestion about audio; if we were to adopt it, it would make a 1D element a "sample".

[1] People are often confused on this point, because in the course of everyday conversation we don't distinguish between the number of pixels on the side of a rectangle (which is a 1D quantity) and the number of pixels inside that rectangle (a 2D quantity). So if I say I have a 10 pixel by 10 pixel image, what I mean is that I have a grid with an area of 100 pixels, with sides measuring 10 pixel-widths by 10 pixel-heights (each a 1D quantity of length). If that looks awkward and tiresomely pedantic to you, well, that's why we just say pixels and let the details be implied.

If you're still skeptical, consider for instance that voxels are more clearly a unit of volume (think Minecraft blocks), and that pixels are obtained by subdividing a rectangle. Another useful way to think about it might be by replacing "pixels" with "dominos" and imagining making grids out of dominos, pixels can be tricky since you can't see their area yourself.

“1D image” is the technical term used in graphics programming for a single row buffer of width * bytes per pixel.

Am I so out of touch? No, it's the children who are wrong!

But in all seriousness, call it what you want, I happen to enjoy this minutia but understand many people see it as an impediment to clear communication. If you're working in the unusual contexts where the difference matters you probably know.

Still, fixed size pixels arranged in a line, where each pixel's position is described by a single coordinate, could be called a 1-dimensional arrangement of pixels. You could do the same with voxels, or corn fields of 1 ha each, or hypercubes.

Yeah, if you have an array of pixels or hypercubes, you can think of it as a 1D array and leave the details of what's contained below your barrier of abstraction. And that's a useful mental model much of the time.

But I would still argue that an array of pixels doesn't represent a 1D image. If a 2D image associates areas with color or intensity values, a 1D image would associate intervals with color or intensity values (since intervals are the measure of 1D space which is analogous to areas in 2D space). In my mind, those are different data structures, but the difference is pretty nuanced and I would understand if people felt I was splitting hairs.

Given the question, "what's a 1D image?" I'd argue this is the more complete answer. But if we were to ask that question in the context of a real world problem, yours is likely to be the more useful answer.

> Pixels are a unit of area like an acre or square meter.

Some computer graphics experts don't agree with this: http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf

I've skimmed the article, it's gunnuh take me a while to read it so that I can respond properly. In case I don't get around to it, thanks for the article. It's an interesting perspective.

1D arrays are usually used for 2D images

An audio file.

Scan lines?

A barcode.

Reply on news.ycombinator.com