MakerBot demos 3D object scanner that fits on your desk MakerBot, producer of the Replicator line of desktop 3D printers, took to the stage at the SXSW Interactive conference in Austin, Texas on Friday to demo a prototype of its first desktop 3D scanner. While 3D printing may be no big deal these days – numerous affordable models are available and enthusiasts are using them to …
for occasionally incorporating 3D models of real world objects in games. Quite the time saver, and it would be nice to see a house in, say, Fallout 3 that contained something other than the usual toaster, lunchbox, Abraxo cleaner, motorcycle handbrake, tin can, medical brace, etc.
Then again having 1 model of a garden gnome take up more space than all the game graphics, music, code and cut scenes might not be so good.
And as for time saved - having a refresh rate of 5fps (on a kick-ass system) might be seen as a bit of a drawback too.
Game models are not low poly and limited in number because they cant be made. Its cos they cant be processed/stored in a timely manner.
quite right, and before you know it we are back to the usual "toaster, lunchbox, Abraxo cleaner, motorcycle handbrake, tin can, medical brace, etc."
I have the most fantastic scenery model of a strip about 1km wide by 27 km long, a reasonably large playing area you might think.
12GB of raw data - just a dtm and trees, buildings roads etc
And my machine with 16 gig of ram, an i7 chip and quadro 3300 graphics card, can manage a few fps - unless i skin the surface and project high res images onto the data, in which case i get about 0.5fps.
The guys who make games for a living invest a massive amount of time and effort into squeezing things down, and doing smart things with textures to the point that the hardware can deal with it and still give a reasonable performance.
every vertex, every poly you add to that slows things down exponentially - more points makes more collision tests, and more points makes each test more complex.
and forget about algorithmic poly reduction - with the knob to the left it accomplished pretty much bugger all and with the knob to the right it ruins the models :-) for a first pass on a model, maybe, but after that takes judgement, not just combining polys with a surface normal within x degrees of one another.
have a play with 3ds max for a few years, then you get some appreciation on how incredibly detailed and complex the world around us is.
Those knobs _do work_ if you learn how to feed them (partition models, simplify, unpartition). Indeed commercial games developement tools do use tens of millions on triangles for most scenes, and then bake those down to a simplified model and a normal map / parallax map / whatever other fake displacement technique you want to use for "in game use". I can pretty much guarantee that those low detail models are auto generated - no artist wants to hand generate three different level of detail models and keep the three in sync.
> The guys who make games for a living invest a massive amount of time and effort into squeezing things down, and doing smart things with textures to the point that the hardware can deal with it and still give a reasonable performance.
Yep, they invest a lot of time in learning how to use the frickin tools so they don't have to hand roll monotonous boring shitty jobs such as simplifying models. Their time is too valuable for that. Yes they will need some hand touch ups for some parts, but the tools do 95% of the grunt work.
> every vertex, every poly you add to that slows things down exponentially
No. Just no. If the game developer has any common sense this will be log<n> at worst, and most decent algorithms are linear time. You don't think they actually use the on-screen geometry for all of the CPU-side tests do you? Oh dear ...
Somebody mount a Kinect or EyeToy to a turntable and slap together a project on github! Then we can generate the 3D model and the texture at the same time as well.
Joking aside it sounds cool but would probably be impractical for video games as such a scanner would probably make the geometry needlessly complex. /shrug who knows.
What would be even cooler is mounting a scanner like this on a pan tilt zoom mount on a small UAV to fly thru structures and systematically scan entire buildings.
"some early adopter has already tried to stick his dick in this thing."
Since "The Digitizer works by placing an object on a spinning platform" that would involve chopping it off first.
I now have an image of it going around and around on a spike like a mini version of the dubious meat on a kebab shop rotisserie.
Desk top fabricators have been around for a while now and have steadily improved. Fine if you're designing something from scratch but what about copying something?
That's always been the problem and now someone's done something about it.
While the whole spinning table/lights/cameras may remind you of the original John Logie Baird TV system it is fairly simple and probably uses mostly off the shelf parts. The software to calibrate it is likely to be very tricky.
Thumbs up for ingenuity. It's the first affordable coordinate measuring system, but I doubt it will be the last.
Has been around for a while and will generate 3d models from regular photographs. Obviously, laser scanning is going to be much better for precision work and cutting down on the amount of post-processing work (less noise and higher resolution), but I doubt that photos can be totally replaced (within reasonable cost limits) when it comes to surface "texture" mapping (by "texture", I mean in the sense of a colour map rather than an actual texture, obviously).
While it's nice to see this new project, I think it's unnecessarily restrictive. Sure, there are plenty of applications where you just want to scan in a 3d object, so having a controlled shot (such as with a fixed camera and turntable, possibly with a set background for calibration) makes sense there. In fact, these kinds of object scanners have been around for many years. But they can't handle lots of real world scanning tasks that would also be nice, eg, scanning room interiors and larger objects that can't physically fit in the control frame like furniture, vehicles, etc. Being able to track location as you enter an object's interior would also be pretty useful (think of the opening tracking shot in, IIRC, Vertigo, for example--the one where the camera tracks through a sign and into a building).
I think that latter kind of scanning (of larger and enclosing objects) is much more interesting from the point of view of developing new virtual reality and augmented reality applications. It's akin to the shift from still photography to films, with the ability to move around in space and time. Think of robots that can locate obstacles (or goal objects) in a 3d space, or terrain/object mapping based on aerial video recordings, inferring an object's motion relative to other scene elements, or even just as a quick and easy way to knock up quick scenes for first-person shooters (eg, Runtfest map for Quake 3) or 3rd-person interactive puzzle games (modern versions of the old Monkey Island style of game). Digitising small objects is all well and good, but it's really more of a time saver than a game changer, IMO.
Hollywood already uses digitizers to create 3D meshes of sculpted work for monsters, masks and whatnot. Digitizers have allowed the creation of good four pointed polygon based 3D mesh. Scanning has been a poor substitute for digitizers because of the random disorganized polygons they generate. It would be nice if they've discovered a process to create good four point poly meshes with scanning. That would be true progress in recording objects to a computer 3D mesh that's useful. That is what I hope is happening here.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
