At first, I thought this was going to be lame. "Moving it with your hand"? C'mon. I was wrong. This is amazing. I want to have Qi Pan's baby. Wait a minute. I'm a man. Well, I have a daughter. I want her to have Qi Pan's baby. She's studying in London right now. I'll call her up and demand she prostrate herself before him and offer up her body. Thanks to guys like Qi, the future is NOW!

yes it looks like a 3d-scanner, but c'mon that 3D-model has absolutely nothing to do with a useable 3D-model.

Nice idea but the demo clearly shows it has a long way to go before it can actually be used to anything.

It looks like it needs some work still. Notice that it sometimes creates artifacts in the edges. The tower and the main roof have an artifact sticking out from the hand in the picture.

Whoever did this forgot to rotate allow multiple rotation and closed envelopment. Having the hand in the picture interfered with the model learning.

Didn't check out the guys website but this could be pretty helpful for people like me. Whenever i try to create new game objects it's always a pain to do modeling, if this scanner of his can export the model after it scanned (which i can't imagine being to hard if he wrote the scanner program) then this could solve alot of problems for me. There still seem to be a few glitches, but with something like this you should not be expecting perfection, and if you can export the model, then you should easily be able to just reshape the meshes (although you might have to cut them off the model if it just exports the model as one solid mesh.)

Point cloud density is way too low to be usable for any applications. You need a laser scanner, structured light scanner, or a linked multi camera setup if you want to actually create usable 3d models. Sure you can do this, but the results are gonna suck and you'll wish you had called that reverse engineering shop down the street after all.

Yes, this is coming from an applications engineer who does optical metrology.

I wonder what the UV layout looks like, and how that texture is stored in memory. The mesh looks nice and low poly, but I'm not sure if it has crossing / overlapping tris (that would be a nightmare if a modeler wanted to fix it up for an in-game model. This kind of thing might make for a nifty I-phone app or as a Scanner for a DSi game.

guys, seriously, no need to beat up on a proof of concept. Obviously it's not perfect because it's the first of its kind. With a little tweaking to the program you could make the model generation edit-friendly so that game modders can easily import models and edit them manually.

But if this can save you 80-90% of your modelling time on the bulk shapes, and then all you have to do is a little artifact trimming it's totally worth it on a dirt cheap budget. I can even see this being amazingly helpful for creating skinning/mapping patterns (you put markers or color sectiong of the physical model, then once imported it'd be as easy as paint-by-number to reskin your model).

honestly go look at the first ever consumer digital cameras and tell me you could have predicted back then what they would become today.

I think you totally missed the point, Paul. He's doing this WITHOUT ALL YOUR EXPENSIVE SUGGESTIONS! Seriously, with a webcam. A WEBCAM. I do a lot of work in metrology as well (both in development and applications) and I can see how intriguing this concept is, without letting my personal bias and desire to sell my own products cloud my opinion. This is obviously a proof of concept introduction, and doesn't seem to be something to be used in a tediously scale-reliant field like you (and I) probably work in on a regular basis. I can see where something like this would work great in graphics arts avenues, like maybe 3D gaming map creation, etc. You know, places where scale can be tapped in with a small hammer, rather than requiring micron precision.

Just give the guy some kudos for a really cool high tech process using (relatively speaking) low tech tools.

Just put the model on a white lazy susan ans spin away

If the only tool you have is a hammer, then every problem looks like a nail

When did I push a product? Do you have any idea who I work for, where I work, or what part of the country I'm in? Nope. I wasn't pushing anything.

I deal with people like this all the time pushing these kinds of things that gets peoples hopes up and turns to dust. Great concept yes, but it just makes my job that much harder when people want this kind of solution and it's totally worthless. Single camera solutions already exist, it's called photogrammetry. This guy hasn't made anything new, it's not a new concept, it's not revolutionary. It may be new to some of you, but for me it's just more of the same stuff that doesn't work.

For the entry level user they are gonna spend 1/4 the time building the model from scratch than if they used this. They will spend so much time cleaning up the data, interpolating points, etc, that they won't save any time. This thing could be modeled in Sketch Up and textured in five minutes and the model quality would be 10 times better.

This isn't anything new. It's using the 2d image with feature point recognition to stitch the images together using a wrap function. There's nothing conceptual here. Add a second camera to capture multiple angles at once for better triangulation, increase the point cloud size, and remove the artifacting issues and you have a decent low end solution for fifth graders.