Hacker News new | past | comments | ask | show | jobs | submit login
Stanford Bunny (wikipedia.org)
121 points by dcminter 5 months ago | hide | past | favorite | 42 comments

I had no idea about Easter when I downloaded and tweaked the mesh model as a collage student. It was too foreign to me as a non-Christian county (Japan) resident. I dreamed to get to see that "special" bunny someday. Now live in U.S. the same-looking bunny is everywhere and I've lost the interest getting one. Still, from time to time I wonder which bunny it exactly was and if I can find one, in the mob.

Tangentially related, I bought a teapot from Melitta on e-bay and am still using it. It's not "that" one, but it shares some of the taste from the brand. The teapot has a piece of uniqueness, unlike the bunny.

It turns out those are still made by the original factory in Germany: https://friesland-porzellan.de/utah-teekanne

That's interesting. The article says that, until recently, the company didn't know that their teapot is so famous.

OP could also be referring to Russell's Teapot, although I think that model stopped production in the 70s.

Also Tangentially related...

What is the bunny in the moon doing?


Hugh Hoppe, from Microsoft, wrote some fundamental progressive-mesh code that he showed at SIGGRAPH '97 using this model. Instead of the code having to choose from different model resolutions (like the original four bunny models), the algorithm would allow DirectX to reduce the vertex count based on the rendered size of the object (1 pixel only needs 1 vertex!) to save compute cycles.

I wonder if geometry compression is still a thing given the performance increases of GPUs. I haven't done 3D driver work since DirectX 5....

Continuous LOD research dropped off dramatically when GPUs got vertex shaders. At that point, having the CPU touching the vertex data at all became more expensive than it could be worth.

Instead, 3 or 4 static LODs have been the way to go. Those static meshes can be individually well optimized. There are surprisingly few good options for auto generating static LODs even today...

Mesh Shaders are opening up whole new opportunities that are only starting to be explored. You get far more general purpose, compute shader like features. But, you have to work in small packets. Funny enough, they are a lot like PlayStation 2 vector unit programming made modern. That design really was ahead of it’s time. But, the fundamental physics advantages of SIMD working on a small chunk of SRAM are as valid as ever.

It's funny you say that because that's effectively what Unreal Engine 5 is doing with it's Nanite technology. Also released in early access today.

I do wonder what the problem was that needed to be solved before we got Nanite. Everyone knew that automatically optimising the geometry for the size on the screen was doable. Maybe it was a storage limitation?

Yeah, geometry compression is definitely still a thing. The main reason is that the relative gap between compute speed and memory bandwidth is still growing. It's pretty common now to bump into situations where redundant computation is faster than reading saved results from memory. Compression is perhaps ironically or surprisingly getting more important as GPU perf increases...

For someone who's moderately interested: http://hhoppe.com/lapped_stonebunny_siggraph_cover.html

Glad to see he's still "the mesh guy", wondering what he's doing at Google. http://hhoppe.com/#publications

Very much still a thing, e.g. for reducing the poly count of a mesh before running it through a procedure to spit out gcode for 3d printing or milling.

Not for GPUs AFAIK but still related: https://www.blosc.org/pages/blosc-in-depth/

> Hugh Hoppe

Hugues (though most people pronounce it fairly close to “Hugh’s”, so maybe that’s what stuck for you?)

It pleases me that someone named Hoppe picked a rabbit for the demo.

Ha! In 24 years I never made that connection.

Man, does that thing bring back memories. I did my PhD in computer graphics (though I no longer work in that field) -- I've written several papers and read more than I can count. That bunny is iconic and still shows up in some absurd number, although it's considered a trivially simple model now. Even so, it can find ways to bite you; there are holes in the base can cause problems with some algorithms (sure you can find watertight versions... but that's cheating ;)

Fun times.

And its friends, Utah teapot, Cornell box and Susanne, the Blender monkey.

And the tree http://web.stanford.edu/~siegelr/stanford/thetree.html

I know the tree is different, but I mention it because they put out some software, named Dryad, that let you explore the latent space of tree parameters - leaf shape, branch angle, etc. And the default tree was one that resembled the mascot. The software was discontinued and I guess the standard Stanford Tree model never took off.

Edit: I can't find it available for download anywhere, so I put my copy on the Internet Archive. https://archive.org/details/stanford-dryad It will take some tweaking to get it to run on newer Windows versions. You can see the Stanford Tree model right in the center of the screenshot on the right here: https://news.stanford.edu/news/2008/january9/dryad-010908.ht...

The Utah teapot, or the ‘teapotahedron’ as it is sometimes termed.

There is also Armadillo and Happy Buddha from Stanford.

(Posting top-level so it isn’t missed several layers down)

For folks that are interested in this kind of thing, Frank Crow’s 1987 article “The Origins of the Teapot” [1] has some great stuff in it (including a small scan of Martin Newell’s graph paper for calculating the points).

[1] https://www.computer.org/csdl/magazine/cg/1987/01/mcg1987010...

I think it’s sad that Lenna had to go, with vague arguments of sexism. She’s a photo model. Most photo models are women, because that’s the reality we live in. I would have thought it cute that my (tiny) country is represented in these contexts, but no, she was a beautiful woman and that’s sexist to use as a test subject.

Without passing any judgement, I think it's important to note that she is a nude photo model. The image is a cropped Playboy centerfold.

Indeed, and I also think it's important to note that it's a picture from the 70's, an era known for relaxed sexual morals, and particularly in Sweden.

The Lenna picture is not something I care deeply about, but the overall trend of post-hoc censoring cultural expressions because some people MAY find them distasteful is letting the bullies win.

I wonder how much of the push against Lenna was veiled jealousy.

Ive always used benchy the boat.


I wonder what rabbits think of their representation in human media.

better than that damned teapot.

The teapot was actually designed with Bezier surfaces. That makes it a whole different animal.

The teapot is a design, the bunny is a scan. That is the most important difference.

What do you mean exactly? The teapot was "scanned" from a physical model, not designed from scratch. I'm quoting "scanned" because it was a manual process, and it'd be fair to say there was a little bit of design involved. But the primary distinction you made here between design and scan isn't really accurate without more qualification or explanation. You can see pictures of the original physical teapot model and the scan data (drawings) on the WP page https://en.wikipedia.org/wiki/Utah_teapot

Perhaps design is not the right word, it’s manually modeled. When you are using the model it doesn’t really matter whether there is a physics reference or not, the data from a manually built model is just pretty different from a scan. Like a drawing vs a photo.

Yeah manually modeled from reference seems like a perfectly fine summary. ‘Drawing’ as a one word summary still tends to imply a pure creative process and not a reproduction process. But, the article @boulos shared is the best description, and the article does describe it as a “sketch” next to a picture of the hand-plot on graph paper.

The page you link to even talks about it being a model rather than a scan:

> It is a mathematical model of an ordinary Melitta-brand teapot

> The teapot model was created in 1975 by early computer graphics researcher Martin Newell, a member of the pioneering graphics program at the University of Utah.[3] It was one of the first to be modeled (using bézier curves) rather than precisely measured.

Yes that's correct. It wasn't scanned in the typical meaning of "scan" at the time, by digitizing the vertices of a triangle mesh. But it also wasn't an original design from scratch - which is what calling it a "design" seems to imply. The Bezier curves were plotted out to match the reference model, the Melitta teapot, so it was in a sense, a manual eyeballed scan. Calling it a "design" as something in opposition to a scan just seems a bit misleading. There is some overlap of those two terms with the Utah Teapot. The shape wasn't an original design, just the topology and representation were, and the digitizing process was manual and approximate, right?

Right. Both the teapot and Ivan’s VW are “manual scans”. The teapot even got “accidentally” stretched by Jim Blinn and most copies use that one.

In searching for that evidence (Wikipedia says citation needed!), I rediscovered Frank Crow’s great write up from 1987 (which includes a scan of the graph paper!):




the teapot was done 2 decades earlier. much more impressive technologically. think about trying to display pixelated graphics in 3d on 70s tech. mind blown.

I'm sure I ate one of these as a child.

The Lindt bunny does look similar but I guess it's lower resolution ;-)

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact