In this experiment I've trained (pre-wired) a Vision Network to recognize 26 fractal type patterns, the network resides on the Numenta server...
http://www.numenta.com/vision/webservices-embed.php?startpage=netinfer&ni...
It receives a little 204 X 204 pixel greyscale image from the iPhone and returns 26 category values indicating how closely the submitted image matches each category.
The iPhone does nothing more than use the top 3 categories (not the values) as the address to the image memory text.
The iPhone is trained to recognize images the way a child would learn, show it something and tell it what it is. It has a photographic memory in that it only needs to be told once, if you show it the same image again it will recognize it. It also has the ability to generalize and you can teach it many views of something. Even in this very simplistic form it can learn thousands of images and there is no searching involved. The image, filtered through the fractal patterns, creates the address (fires the Bill Clinton neuron ;)
Learn more about HTM and the Vision Toolkit at http://www.numenta.com and search iTunes for Tech Center Labs for free iPhone apps using the Vision Toolkit.
Gary, http://www.iTalentProductions.com
Link to this comment:
All Comments (0)