Universal 3D File Format In The Works

telstar writes "The Register is reporting that more than 30 companies are working together to define a new file format intended to serve as a universal 3D file format. The new file format will be named the 'Universal 3D Format', or U3D. According to the article, they hope to make the new format as standard as MP3 has become for audio, and JPEG has become for 2D images. Interesting that they would choose two lossy media formats as models for comparison."

There's a page for this on the ECMA site...

[tcopeland]

...right here [ecma-international.org].

There's also a separate 3DIF [3dif.org] site.

.3ds

[TheRealMindChild]

Ok, so its ASSOCIATED with some random application (3d studio max), but ANYTHING that does 3d will read/write to a .3ds file, if they take themselves seriously. Whats wrong with that?

JPEG patent is bullshit

[tepples]

You have a point with MP3, but the author of BurnAllGIFs.org [burnallgifs.org] seems to think the JPEG patent wouldn't stand up in a court of law.

Control mesh noise with a slider

[tepples]

[coordinate noise would] create a jittery effect

In skeletal animation, noise in the mesh would move more or less rigidly with each bone, creating a bit of roughness but no jitter. In non-skeletal animation, one could move a slider to increase the precision with which the animation tool stores coordinates. Remember that even 64-bit floating point isn't perfect.

Comparison to formats...

[ItMustBeEsoteric]

Most likely has to do with the fact that .mp3 and .jpeg are ubiquitous more so than anything. If they said, for example, .FLAC, they would have confuzlled the hell out of a good chunk of people.

Re:Really bad examples to pick...

[Anonymous Coward]

The GIF patent has expired. You are allowed to freely do things with GIF now.

Thank you for being up to date on all of your legal technical issues.

Re:Standards

[breon.halling]

According to this [ucl.ac.uk], it was Grace Hopper [sdsc.edu].

Geometry Images

[duckpoopy]

Conformal surface parameterization [ufl.edu] allows you generate a geometry image from an arbitrary mesh. The geometry image is a parameterization of the mesh on a uniform grid where the (r,g,b) coordinates are considered to be (x,y,z) spatial coordinates. You can now use the image format of your choice, lossy or not.

Re:Really bad examples to pick...

[Anonymous Coward]

Only in the US. Other countries are still hanging on.

"The U.S. LZW patent expire[d] June 20, 2003, the counterpart Canadian patent expires July 7, 2004, the counterpart patents in the United Kingdom, France, Germany and Italy expire June 18, 2004, and the Japanese counterpart patents expire June 20, 2004."

Re:What about VRML

[Urban Garlic]

VRML, as I understand it, is not quite dead yet. There's still an active W3G working group working on "x3d", which is the successor to VRML, and more than a few scientific visualization packages can export to, and read, VRML, including SGI's OpenInventor.

What's bad about VRML was that the VRML '97 spec was too damn complicated (IMHO), and a few years later, the really good free browser (CosmoPlayer) got sold off by SGI, and after changing hands several times, apparently disappeared from the face of the Earth. There are other browsers, but they don't plug in to browsers as easily.

The other problem I heard people complain about (but was not a problem for me) was the "JavaScript" problem -- people on comp.lang.vrml didn't like that their web VRML was human-readable and stealable. CNN used to have the occasional VRML model on their site for interesting things, but switched a while ago to something called Cult3D, which appears to be binary, and to have pricey development tools -- I don't know if the format is actually proprietary, but it wouldn't surprise me.

Of course, the *real* reason it died was because I learned it...

Why VRML sucked

[sjbe]

VRML had a bunch of problems.

• VRML files are huge. VRML was designed to be human readable (a laudable goal) but this meant that a VRML of any complexity made an enormous file. I used to use VRML files out of ProEngineer in a 3D simulation package and the smallest files were usually about 20 megabytes and it went up to 200 at times.
  
• VRML lacked the precision to be useful as a CAD quality 3D data interchange format. Not that any of the alternatives are great (IGES) but VRML didn't solve the very real problem of interchange between incompatible 3D modeling systems.
  
• VRML was designed (partly) with the idea of moving the web to 3D, but this isn't very useful for real world interfaces. Even where 3D interfaces might be useful (rare) there isn't the infrastructure (bandwidth, control systems, etc) to make it useful
  
• Unrealistic (read poor) image quality. You will never see a VRML file that looks anything close to as good as a modern 3D shooter.
  

Basically VRML wasn't designed to scratch a real itch, just a theoretical one. It was just a neat idea that was designed by committee, with predictable results.

Missing Members, universal format?

[PenguinOpus]

At least according to the reports I could find, Alias and Autodesk are not members/founders of 3DIF. I would be very surprised if a "universal" 3D format is created without their help.

The problem is that there are many diverse needs/users for 3D data and the data is so large and/or hard-to-render that simplifying it to a single format that meets the needs of CAD, film animation, sci-vis, game animation, 3D web content, and GIS is not reasonable.

On the other hand, VRML may have been too early, but a poly+texture+simplebehaviours format that was well supported by all applications would be a good thing to (re)create.

Re:How is it going to handle programmable shaders?

[Mithrandir]

Unless they have done some serious modifications to it from when we were working with them on it 12 months ago, the architecture prohibits the use of any video hardware accelerated capabilities - let alone programmable shaders. The format implicitly requires CPU utilitisation all the way through until you hit the rasterisation stage. If you want to see why, do some research into the Modifier Chain architecture part of the spec. It's a great concept, but totally in appropriate for hardware acceleration.

Re:I was excited for a moment...

[Mithrandir]

There is one in the works - There's a CAD profile of the X3D standard that is currently being worked on by the Web3d Consortium [web3d.org]. The primary purpose is for data interchange at the DCC tool level so that CAD content can be taken and repurposed for tasks like training manuals, applications and online display. Every major CAD vendor as well as all the other tool vendors (Autodesk, Alias etc etc) are involved. The standard is underway and the first draft to be added to the ISO process should be ready in the next month or so.

Re:already an iso 3d standard

[drfrog]

correct url is web3d.org

I like the idea.

[moroderzone]

I do research in computer graphics. Finding 3d models is a pain. There are so many formats, and converting between formats is no fun. Sometimes parts of the model get messed up. For example the surfaces normals come out backwards or the material properties disappear. I think if they release a nice file format, and a nice and fast API for loading and saving files i would check it out. And if there were a lot of nice 3d models available in that format, i would be all for it.

Re:What about VRML

[jherekc]

To the best of my knowledge, there's no good way to transfer animated scenes between programs (if I'm wrong on this, PLEASE correct me)

dammit, has no-one on slashdot heard of RIB?

JPEG is not a lossy format

[alphakappa]

JPEG is not *necessarily* lossy. The JPEG specification allows for both lossy and lossless compression.

In common parlance, however, JPEG refers to the *JPEG baseline algorithm* which is lossy (but allows you to define the amount of loss). Note that even though you can create images that are visually lossless, baseline JPEG can never produce truly (mathematically) lossless compression. (no, not even if you set quality=100)

If you want lossless JPEG compression, there's the standard called *lossless JPEG* (LJPEG) which doesn't provide a high degree of compression though. There's also *JPEG-LS* which is another JPEG standard which provides for lossless compression.

If that's not enough JPEG for you, there's the new standard called *JPEG 2000* which allows a host of features such as the ability to choose between lossy and lossless compression, progressive transmission etc.

So calling JPEG lossy is true only if you are referring to baseline JPEG.

hopefully they get it right

[Sludge]

Do something cool, and THEN move to standardize it. This was the fundamental problem with VRML. Standards locked it down far before it was useful.

Overambitious?

[Dr. Mu]

The goals of this project are either overambitious or overstated. Each of the 3D file formats in current use has its plusses and minuses, depending on what it's used for. For example, VRML works fine for dynamic 3D visualation, but I wouldn't want to fabricate anything from it. Likewise, STL is optimized for stereolithography, but IGES is de regeur for final manufacturing. A single, universal format would have to include a lot more data than any one single application could ever need.

Perhaps the best approach is a pseudo file format with plug in codecs, like Microsoft uses for its video playback.

.u3d already in use!

[buhatkj]

the .u3d file extension is already used for a proprietary format for a neat little app called uinwrap3d, which a lot of modders use to make skins for custom game models. prolly no big deal but i just thought that was mildly relavent
maybe not...

Re:Lossiness? No, try patents

[fredmosby]

It doesn't really matter because their format will not be lossy. 3D files aren't like graphics or MP3's, the only reason a point or polygon exists in a 3D file is because an artist told the computer to put it there. In an uncompressed bitmap a white page would take as much space as a photograph of a house. An empty 3-D file would take much less space than a 3-D file containing a model house.

Already are standards...

[malfaetor]

I work in a position where I get to mess around with transferring geometry between CAD applications almost daily. Thankfully for me, there are already a couple of standards for 3D Data Transfer.

High-quality 3D CAD programs *should* already be able to import the Initial Graphics Exchange Specification (IGES) [nist.gov]. IGES has absolutely been the de-facto standard for the past 10 years on data transfer between CAD packages. It handles surfaces, 2D drawings, 3D wireframe, as well as solids (although it didn't originally support solids). Unfortunately, some CAD software manufacturers *cough*AutoCAD*cough* force the consumer to buy an additional license to handle it. They want everyone to use their proprietary Drawing Exchange Format (DXF)

Some CAD packages had attempted to go to the solid model transfer format STEP (Standard for the Exchange of Product Model Data), but few have been quick to adopt it. STEP works extremely well between CAD packages that do support the format. It'll even build assembly heirarchies for the user as necessary. Unfortunately, STEP doesn't handle parametric models (models driven by dimensions, instead of the other way around).

That said, there are still some downfalls of all flavors of the current intermediate transfer formats. I look forward to the day when I don't have to worry about what format a given CAD package uses, and how they interact with each other.

-Malfaetor

Reviled did I live, said I, as evil I did deliver.

Re:.u3d already in use!

[Anonymous Coward]

pffft... hardly.

u3d is the standard Unreal engine's 3D format.

Dear XML ...

[jefu]

XML models could be large, very large even, but XML in general compresses well. Furthermore, it is well suited for hierarchical model building so would probably do well in 3D models.

And it is quite possible to specify an XML format which would serve as the "canonical" format for a model and specify one or more "shorthand" formats that encode the XML format more concisely with rules for transforming one to the other.

XML has several advantages - it is text based so can be easily edited by humans when necessary and there are XML editors that can simplify the process. It has many standard tools and toolchains and XSLT is maturing nicely as a transformation engine which provides for lots of additional capabilities (and don't forget XQuery, and native XML databases). Finally, it is by nature extensible, allowing for different ways to put in comments, add in vendor specific extensions that are easily ignored by other vendors (or used when possible), provide for upgrade paths and the like.

On the whole, using XML has disadvantages, but advantages as well. Given a choice between large XML and some smaller but quasi-proprietary binary format, I'll take XML every time.

Re:VRML is now X3D!

[Mithrandir]

You're partially correct. X3D added a XML format, it didn't replace the existing one with it. As such it has grown a lot so now you can express the same concepts in 3 different file formats - the origial Inventor-style curly brackets, XML and a pure binary form.

As for the other technologies, they're included in the X3D standard. The only ones taht aren't are progammable shaders and 3D texturing support. Both of those are currently going through the standardisationn process working groups within the Web3d consortium.

Only a part of X3D is incorporated into MPEG IV. There's a lot they didn't take - most specifically all the extensibility that X3D allows. It's a single fixed profile of functionality. It's a rather cut-down version if anything.

Re:VRML isn't just a description language

[Cochonou]

VRML files are huge. VRML was designed to be human readable (a laudable goal) but this meant that a VRML of any complexity made an enormous file. I used to use VRML files out of ProEngineer in a 3D simulation package and the smallest files were usually about 20 megabytes and it went up to 200 at times.

VRML files are designed to be human readable because VRML isn't just a mere 3D description language, but also a programming language.
The very big advantage of VRML/X3D for designing virtual worlds is that you can not only design objects with VRML, but also define the interactions between them directly in the VRML source.
Moreover, the concept of scene graph (the 3D scene is a tree, if you affect an object its children are affected) which was by popularized by VRML has proven to be quite effective for developping virual worlds, and has been for example adopted in newer technologies like Java 3D.