<< return to Pixycam.com

Colour codes and then shape codes?

Hi, this talk of supporting facial recognition (at some point in the future) is confusing me - isn’t the unique feature of Pixy the fact that it can track multiple colours (and colour codes) at high speed? So how does that apply to facial recognition, which uses shapes and lines, as opposed to colour. Aren’t they very different algorithms?

The ability to also track shape codes (say a circle overlapping a square or something) would probably have more practical applications than tracking colours, especially if it could do it really fast. For instance tracking multiple QR codes (http://en.wikipedia.org/wiki/QR_code) in real-time would be pretty handy.

QR codes is an interesting idea for sure.

That’s face tracking not face recognition - two different things. Face tracking is much faster. It is very useful for robots that interact with people. There are lots of commercial products like digital cameras that have face detection and tracking built in to do things like set focus or exposure settings for people’s faces.

Hi Paul,

Like Alyssa said, there are many applications that face tracking can be applied to, which is why we’re going to include it in Pixy. To clarify, the object tracking algorithm in Pixy is not what we’ll use for tracking faces. “This”:http://cmucam.org/projects/cmucam5/wiki/Can_Pixy_do_face_tracking_face_detection_or_facial_recognition page in the wiki provides more information on what we’re thinking for facial tracking.

Scott

I’d like to see shape tracking, too. Not for faces but for tracking an object by color + shape. Specifically I’d like to be able to sense alignment of the edges of objects so that when they’re overlapping (edges line up), that is detected.

Thanks!