Nothing ventured, nothing gained

Riya just released Like.com, a visual search engine geared towards women and their collective billions.

Imagine this, you’re walking down ASU Palm Walk and of all the people in the world, you run into Jessica Alba who is sporting a pair of black shoes. Since they looked incredible on Jessica, the first thought that runs through your mind is to get your girlfriend a pair just like it. The problem is, you lack the basic knowledge required to purchase female foot accessories. So you do the next best thing: you take a picture of her shoes in the hopes that one day, someone will invent a search engine where you can search for shoes, visually. Then one day, you meet a guy who runs a Silicon Valley start up called Like.com. Like.com let’s you search for images, visually. According to TechCrunch, Like.com creates a “visual signature” for the query image, where the signature is a mathematical representation of the image using 10,000 variables. If enough variables are identical, Like.com decides the images are similar. Right now, the site seems to be geared towards female accessories like shoes, bags, watches, etc. Some interesting tidbits from a data mining perspective:

1. I remember Riya when it’s facial recognition software caught headlines back in early 2006; there were rumors of being bought out by Google. Good thing they didn’t because the work with Like.com looks promising.
2. It turns out that doing facial recognition was too complicated, at least according to user feedback, so they changed the game by going after a different market: women and non-geeks.
3. The engine needs to work on high-resolution versions of the image.
4. The software deconstructs these images into approximately 10,000 data points (or, I’m guessing attributes), which include:
5. Characteristics like color, texture, material, shape, pattern, brand, style, and blingness, etc.
6. To process the jewelry set takes 20GB of RAM.
7. Shoes and bags are the easiest items to search for. Which is just as well since they are probably the easiest to shop for.

Now, you can take that picture of Jessica’s shoes and start punching in words that describe it. Black. Shoes :-). Ok, ok, Black, open, toe, polka, dot, shoes. Viola.

I showed my girlfriend the website:

Michelle: Type in Manolo Blahnik Me: Umm, [as I type in “M-a-n-a-l-o B-l-a-n-i-k”] and how do you spell that [Hits Enter]? Oh cool, it made a spelling suggestion, look “Manolo Blahnik.”

Pretty cool stuff!