Giving computers the gift of sight

Leading researchers at the Courant Institute of Mathematical Sciences and the Massachusetts Institute of Technology are developing computer sight technologies that will allow machines to observe and interpret visual information the way humans do.

Rob Fergus, assistant professor of computer science at Courant, is one of the researchers working to develop a system that not only records visual information, but also has the capability to identify images based on color values and positions.

“The human eye is just a sensor. It’s the brain that interprets and organizes the information we see,” Fergus said.

Not every image has a corresponding label, which greatly inhibits the computer’s ability to search for visual media. The new system would search all pictures online and automatically tag them.

The emerging field of computer sight would create a database that knows what a car, an animal or a person looks like, and the computer would be able to identify these objects in an image without a text label.

“There’s the problem of retrieval and also the task of recognition [of images],” Fergus said. “We’re working on methods to make both of these more effective.”

When uploading digital pictures to their computers, photographers would only have to describe a person once to the computer. A computer sight program could then label all other pictures of that person, as well as identify any other objects the user specifies.

The practical applications of this technology are numerous. Airport security terminals, for instance, would be able to quickly scan all boarding passengers and apprehend known terrorists.

In the medical field, an automated program could be used to sift through thousands of X-rays, MRIs and CAT scans to determine whether or not patients have bone fractures, tumors or glaucoma.

Computer sight technologies could also potentially alleviate vision problems. Developing these technologies requires an intensive study of the brain’s ability to process visual signals, which allows researchers to learn more about sight itself.

With computers growing increasingly intelligent, some may fear science fiction becoming reality. SCPS sophomore Katherine Lau is worried that computers might become too smart.

“I think that it’s fine, as long as they don’t violate human codes of decency,” Lau said.

Fergus dismissed the idea that computers would naturally overstep their boundaries.

“All technology can be used for good or evil. You could use it for the battlefield or to help the sick,” Fergus said.

Stern freshman Lov Patel is optimistic about the new technology.

“I think it would be unwise to halt the progress of technology just because human-like robots seem threatening to some people. Robots are still representative of human advancements, and I think they can really improve human lives,” Patel said.

Regardless of how advanced these technologies become, Fergus believes there isn’t any reason to be concerned about a robot takeover.

“I wouldn’t worry too much about that kind of thing,” he said.

Kaivan Sattar is a contributing writer. E-mail him at university@nyunews.com.

Comments

1 comments

Your name

Tue Sep 30 2008 08:43

This techology would really be a huge wonder.We should be all working on these tyoe of projects.

Fahad Farooqi
BSCS(Software Engineer) GCU, PAkistan.

Add comment


I am not posting spam. I understand posting spam or other comments that are unrelated to this article will cause my comment to be flagged for deletion and possibly cause my IP address to be permanently banned from this server.

Enter the characters in the image above: