This Software Flags "Inappropriate" Google Glass Photos

Don't want Glass snapping pictures while you're in a public restroom? Meet PlaceAvoider, a new image-detection system being developed at Indiana University.

Google Glass, which just got 200% less dorky looking now that it affixes to prescription lenses, is still at a point where it raises more questions than it answers. And one of those questions is: When is Glass inappropriate to wear?

The answer might seem obvious to you, Person with Common Decency, but for others who might have a tougher time picking up obvious social cues, the etiquette of the future might not be so straightforward.

Enter a new software technology called PlaceAvoider, which is currently being developed by imaging experts at Indiana University. Its goal is to take out a lot of the guesswork for would-be photographers by automatically blacklisting images taken in inappropriate settings—accidental or otherwise. Apu Kapadia, who co-created the system, tells MIT Technology Review that the project aims "to help people exploit these applications to the full by providing them with a way to share safely."

How does it work? Well, it doesn't totally—at least not yet. (You can read the paper in PDF form here.) But the general idea is that PlaceAvoider uses algorithms to gauge your location using a system called scale-invariant feature transform (SIFT). That's a fancy name for something that automatically susses out where you are based on a photo. Certain rooms can be blacklisted by the user. For example, the imaging software can be trained to recognize when you're inside a public bathroom versus when you're out walking your dog. If you use Glass to snap a photo in a place deemed sensitive—restrooms, movie theaters, board meetings, etc.—that image is flagged or quarantined.

In reality, it isn't so different from some of the facial-recognition systems we've already seen, only instead of scanning for human facial features, PlaceAvoider's algorithms scan for toilets. And so far, at least, the technology's future looks promising: The prototype made it through the initial round of image guesswork with 89.8% accuracy.

[Image: Wikipedia]

Add New Comment

0 Comments