next »

[Craig Stern]

I just found

on Youtube; there's a program out there where you can load a series of photographs and have it generate a cloud of points, from which you can generate 3D meshes. I think it's kind of amazing:





Just curious: has anybody tried this technique before?

[Bree]

Apart from the fact that you'll need over 400 photos to make it work, that is absolutely incredible.

[Jrsquee]

time to try this

[Halcyon]

Do show the results.

Very awesome indeed, though I wonder how time consuming it is.

[agj]

That is quite something. I wonder, can the software read the correct position of the pictures automatically? If so, it'd be so easy to just grab a camera and use 'continuous shooting' while circling around the subject.

[havchr]

So.. build a stick with multiple cam-corders pointing in different directions. Record while you walking around for a bit. Pipe frames from video into photosynth.
Out comes a 3d scan of the environment you walked around in.
Kinda nice

[Eclipse]

well you'll end up having an unusable 3d mesh.. or at least usable just like the Stanford bunny, but it's a nice piece of software indeed

[Zaknafein]

I'm kinda bummed out to see that they didn't use texture information! If they could extract depth, certainly they can generate a texture out of all the photographic information.
It wouldn't be easy (different resolutions and projection planes for each photo), but it has to be possible, and it would be awesome...