Giving Machines Sight: A “Smart” Camera That Can Learn

The idea of enabling machines to see is one that has far-reaching applications across a whole range of industries and fields—from manufacturing to robotics to military to exploratory. It’s an exciting and slightly unnerving prospect, conjuring up images of sentient machines and autonomous weapons. Someone who’s helping to push this technology onwards is Zdenek Kalal, a research student at University of Surrey, UK, whose work explores real-time vision algorithms. His recent project is called Predator and it’s a smart camera that can not only track and detect objects in real-time, but can also learn from its errors.

Kalal explains:

TLD (aka Predator) is an algorithm for tracking unknown objects in unconstrained video streams. The object of interest is defined by a bounding box in a single frame. TLD simultaneously tracks the object, learns its appearance and detects it whenever it appears in the video. The result is a real-time tracking that typically improves over time.

Pair Kalal’s project with this adorably creepy Kinect object recognition demo from a few months back, and suddenly sentient machines don’t seem like such a distant future.

The applications for this technology are fairly huge from human computer interfaces to augmented reality, video games, motion capture, autonomous vehicles, maybe even space exploration. Along with the more Orwellian connotations like improved surveillance. And who knows, maybe one day we’ll end up with cyborgs who can analyze the world just like the Terminator.

[via Bag of Brains]