Connecting People of Like Interests

[UNDER CONSTRUCTION.
Also, make blog for: Connect search history to Google maps; show keyword search similarities by area so people can see where other people like them congregate.  Idk, just an idea].

Overview:  Technology such as Google Glass is going to let us see reality through a lens augmented by the internet.  In time, I can see this leading to a world where we have the capability to see through the eyes of others.  The topic of this blog is on how we can use Google Glass to help connect people of like interests.

Auras:  Basic Idea

Once Google Glass is more common place, we could create a Common Interests website where users can list things they are interested in connecting with other people on.  These can be things like books they've read, movies they like, political affiliations, etc.  These can also be mined from things like Facebook accounts if users want to link them.

To illustrate the concept, consider the following example.

Assume that Person A has the following set of likes/dislikes.

Likes:  Puppies, Skiing, Dancing
Loves:  Tacos, Burritos
Dislikes:  Kitties, Spinach
Hates:  Sushi, Rap Music


Assume that Person B has the following set of likes/dislikes.

Likes:  Puppies, Skiing, Video Games
Loves:  Tacos, Cheeseburgers
Dislikes:  Kitties, Trigonometry
Hates:  Sushi, Country Music

A caricature of what Person A would see when viewing Person B (and vice versa) is depicted below.

Figure 1.  What would be seen through Google Glass when the people in the example view each other.

The visual is comprised of the overlap between the likes and dislikes of the two individuals, so that they only see what they have in common.  This could additionally be used to help with dating, as there is nothing preventing incorporating whether or not someone is single and what they are looking for.

Identifying Objects in the Real World

-  We could use 3D barcode labels to identify objects we encounter in the real world; these objects can be things like the faces of individuals.  Input a 2D projection of the 3D barcode, check to see if that projection is a possible projection of the 3D label, if it passes the check, link the information for the 3D label to the 2D representation of the object as depicted on the screen.

Problem:  You need 3 such (random) projections to uniquely identify the object

Solution:  OK, so just take 3 random projections before doing the check; not a big deal.


Problem: Both the 3D representation and 2D representation are at the same scale (i.e. one isn't how the object would look a mile away, and the other an inch away?)

Solution:  We would need to construct the 3D image from light rays hitting the 3D label being examined.  But what would define the boundaries of the label?  We could do community detection on groups of adjacent pixels.  Make a network where each pixel is connected to each adjacent pixel, and treat it as a community detection problem.

As an example, consider the following 3 pictures of a hat.

We can detect shadows by cataloging light sources and the types of light they produce.  We can check what kind of light source we are looking at by aggregating clusters of pixels based on both spacial location and color value.  We can do this by treating each pixel as a node in a network that is connected to adjacent pixels.  These edges should be weighted according to a similarity measure based on color values, and we can pull out aggregates using a network oriented community detection algorithm.  We should be able to detect shadows by checking variations in color values within communities.  Light color would be indicated by the variations in color+intensity space along sequences of adjacent pixels.

Once we know what color the light source(s) is(are), we should be able to reconstruct their location in physical space by using the configuration of shadows in the picture.  Once we have done that, we should then be able to construct a 3D representation of the image we are looking at.