Teaching an old dog a new 3-D trick
A newly developed search engine could usher in a new age of exploration where consumers can find what they want through drawings.
Search engines based solely on word entries may soon get a new feature that allows users to draw the object they need to find. This is the hope of Karthik Ramani, a Purdue University professor who has created a system for sifting through computer-designed industrial parts.
This new system was created for use in business markets for engineers to easily find pieces that may already exist instead of fabricating parts on the spot. This has been a problem in the past because, although parts may exist, there is no way for the engineers in the field to know this.
By utilizing the search engine, the workers are able to draw the certain part they are looking for to see if there is something that fits their needs.
After they submit their queries, the computer returns a number of possible matches to their drawings. If the users are not satisfied with the findings, they can manipulate a virtual skeleton of the drawing in hopes of getting different results.
The use of a virtual skeleton is at the heart of how the system functions. To create the skeleton, the computer uses millions of tiny measurements of space that are similar to pixels on a digital picture. These voxels, as the measurements are called, are the smallest division of a three-dimensional space or image that could be likened to a small cube.
Ramani’s program creates voxel-based objects by converting computer-aided designs (CADs) created by engineers and architects. These CADs are those sketches that were formally drawn with pencil and paper, such as plans for a house.
Once the drawings are converted to voxels, the search program is better able to single out components that make up that object. Since the voxel method is based on volume, it will be able to find objects, even if they are mostly hollow.
An example of this would be a coffee mug, which could be distinguished by its handle in spite of the fact that its interior is composed mostly of empty space.
Although the search technology is being developed for business use, it could serve other purposes, such as helping people identify an antique they may have found.
Instead of searching countless hours through text-based systems, the owner could draw a rough sketch to find something that may resemble the object being searched.
While this may not be the intended use for the technology, it could shed some light on the limitless possibilities offered by a visual search engine.
Princeton University researchers already offer a 3-D search engine that allows a visitor to draw an object with his or her mouse and add textual descriptions to find items. This engine,www.cs.princeton.edu/gfx/proj/shape/, gives a good representation of how the new system operates.
The 36,000-object Princeton system returned a number of results for a crudely drawn human head that included a palm tree, a human head, a tree, a pear and a toothbrush holder. It was able to distinguish the drawing well enough to offer a number of possible choices.
Although it’s not perfect, the Princeton system has many of the features of Ramani’s program, including the ability to manipulate the drawing’s “skeleton” to create a new set of possible choices.
Aside from all the proposed uses for the 3-D technology, there are companies, such as Boeing, that have already implemented the search engine into their systems in an effort to reuse parts. The company’s own engineers developed the search method a few years ago.
Mainstream adoption of the 3-D method may still be many years away, but in the meantime, search sites such as Yahoo! and Google offer the ability to look for pictures.
This method may be the next step in the evolution of searching capabilities due to the crude methodology used to find the pictures.
In order to retrieve images, these companies rely on the key words within indexed Web sites. Although this method is somewhat effective, it could evolve into a system whereby the search computers find pictures through a visitor’s sketch.
Speculation aside, according to an Associated Press story, Google’s director of search quality, Peter Norvig, said he is keeping an eye on the newly developed 3-D search engine to see whether it deserves the company’s interest.
Although this technology could eventually make its way to mainstream search engines, it will not mean the demise of solely text-based systems. The reason being that there are still some conceptual topics that could never be expressed through images.
It may, however, give visitors an added tool in their online search arsenal.