Computer Vision Is A Tough Gig
When you are able to use all your senses without difficulty, it can be hard to appreciate what we had to go through as infants to get all those things working; lots of trial and error, and a few scraped knees at least. Imagine for a moment how hard it it for a computer to visually recognize something, like a chair. Remember that "seeing" and "recognizing" are two very different things. The seeing part is easy, TV cameras are small and cheap - you can even use two for stereo vision, no problem.
Now think of all the colors and shapes of something that we generically call "a chair". All those different objects are chairs; how can you even start to make a computer understand that? "Chairs have 4 legs", you may say "that's a good starting point." - okay, so does a sheep. See where I am going with this? Nevertheless, computer object recognition is improving all the time, and such capability is very important if robots are to be able to move around among humans and to be useful at performing tasks "on the fly". The Register has good article that goes over some of the problems and some of the solutions.
|Robby helps Anne Francis with her wardrobe.|