Image Based search-engines (Now search for that elusive watch!)

How many of you have tried to search for that particular "shape", "Design of Watch", "Bridal Embroidery", "Jewelery Design" "Custom Paint-Job" on the net? Like me, all of you have been searching the net based on only one thing: WORDS.
The "Traditional" searches on the net have always relied on "Words" to throw "Likely" results. Search-Engines ultimately look at just words in the form of Tags, met-tags, titles, names, etc to "guess" your results.

In this entry I am referring only to searches for creative & "visual" aspects where you have hit a wall/dead-End. A Google image search for “Tiger” yields many tiger photos – but also returns images of a tiger pear cactus stuck in a tire, a racecar, Tiger Woods, the boxer Dick Tiger, Antarctica, and many others. Why? Today’s large Internet search engines look for images using captions or other text linked to images rather than looking at what is actually in the picture.

Welcome to the new age of Visual-Based / Image-based search engines. This is an entirely different concept from "Image Search Engines". These new-age search engines, will not rely on "mere words" to throw search results.
So, how does this work? Well,the Engine uses Supervised Multiclass Labeling.

Supervised refers to the fact that the users initially train the image labeling system to identify classes of objects, such as “tigers,” “mountains” and “blossoms,” by exposing the system to many different pictures of tigers, grass and blossoms, fruits. The supervised approach allows the system to differentiate between similar visual concepts – such as polar bears, grizzly bears, koala bears. “Multiclass” means that the training process can be repeated for many visual concepts. “Labeling” refers to the process of linking specific features within images directly to words that describe these features.

Schematically, it can be explained as below:

There has been quite a flurry in this field:

Vizseek was initially created to use a photograph, a 2D image, or a 3D model and transform it into a 3D shape & search for relevant matches. Picollator is another example. Even Vodafone has jumped the bandwagon with their Otello. There's alsoTineye

The latest offering in this flooded market is Gazopa from Hitachi. Wonder why an electronics giant would want to launch an image-based search engine. But, I personally, thought it to be the simplest so far.

My verdict is though the initial searches will not give the most relevant results, a few years down the line, when people have "good-heartedly" tagged the images correctly, then your search for "That particular watch I saw at Macy's / Indraprastha (Borivli) / Heera-Panna & clicked its photo" will give you some decent results & leads to the Actual Brands or Knock-offs (whichever you might be interested).

Till then, you are left with simple "words" to "describe" what you search. All the Best.

Do give us a comment after using the mentioned Search-Engines.