Kirt's Cogitations™ #174
Voxels are the way of the future for
search engines designed to mine for 3-dimensional objects across the web. Car parts,
furniture, art collection object, or just about anything represented by a 3-D vector
file format. Researchers at Purdue University have created a method whereby inputting
a 3-dimensional sketch of, say, a football, will result in files containing objects
that have a shape close to that of a football.
For the system to work, cooperative
users must make the 3-D files available for searching. The most likely early adopters
will be manufacturers and distributors with inventories of solid parts, like aircraft,
automotive, appliance, tool, plumbing, and furniture products. Once a standard is
defined, a whole new dimension (pun intended) of resources will be available to
folks looking for parts that can be described by what will become a simple 3-D sketching
Today's search engines perform what can loosely be described as
a 1- or 2-dimensional search for images. Even so, the image pixels themselves are
not actually scanned for content, but relies on file names and descriptive "Alt"
text to clue in the search engine. That is where the voxel comes in. Whereas a pixel
represents a color and a location in an image, a voxel represents a volume (hence
the "V" in voxel) at a given point in the solid object. Since both the presence
and absence of material is represented, a 3-D search engine can tell the difference
between a disk with rounded edges and a toroid that is basically the same shape,
only with a hole in the middle.
Thomas Funkhouser, a Princeton University
professor, has put a 3-D search
engine (sketch applet by
Takeo Igarashi) on the Web that lets the user sketch an object using a computer
mouse, add a text description, then search for similar models in design databases.
Once there, click on the Text & 3-D Sketch link, and play around with the application.
I sketched a hollow bowl and it found bowls and pots without entering any text keywords