#HTE

What Computer Vision Actually Looks Like

Over the years we’ve been conditioned by movies, from the “Terminator” flicks of the ‘80s and '90s to more recently, the “Fast & Furious” franchise’s “God’s Eye,” to have a concept of what computer vision looks like.

But how does the real thing stack up? Take a look at this computer vision system’s take on Times Square:

That’s Google Research’s NASnet, a “state of the art object detection framework.” I was interested to see that it not only picks up vehicles, people, bicycles and traffic lights, but also backpacks, handbags and even drinks cups.

I have mixed feelings on watching this technology progress. Obviously it’s coming no matter what, and I’d like a system that can detect, say, a known terrorist hauling a backpack. On the other hand, if you haven’t yet seen that video of miniature assassin drones–which are basically bullets that can move in any direction and have someone’s name written on them–you need to check it out. If that technology falls into the wrong hands, we’ll be wishing we lived with Terminators instead.

Incidentally, I believe the Daft Punk soundtrack was not added in post-production, but that all computer vision systems spontaneously generate French house music. I’m not a scientist but I know that it’s integral to the computational process.

Which gives me an idea: What would Daft Punk Terminators look like?

“We’re going to funky kill you”



http://www.core77.com/posts/70679/What-Computer-Vision-Actually-Looks-Like