Similarity search introduced by Google

A major step in the direction of combining context and content has been taken by Google. They introduced similarity search. So after you search for images using tradition Google image search, you could then ask for similar images. It will then return all images similar to the one that you select. It appears — have not played with it yet — that it searches for images among the set returned using the meta-data (a subset of context) search commonly used by Google but this time it uses content for similarity and uses context to keep it focused. In their demos it does look good — definitely looks better than Like.com.

Good idea.

5 thoughts on “Similarity search introduced by Google

  1. dinesh vadhia

    Google Lab’s revised similarity image search has garnered wide media coverage. To compare and contrast have a look at http://www.xyggy.com/image.php. Drag and drop ONE or MORE images from the left hand set of images into the search oval. Results will appear on the right hand side.

    Which is cooler and more useful? Xyggy or Google image search?

    Dinesh

  2. Mike

    It is very cool. It makes life a ton easier for some of the projects I am working on. I also agree it is much better than Like.com

  3. Jieleah

    I need to agree that this Google Similar Images is a good idea. It will lessen the time for looking images because it already provides related information. Well, you said it right more better than live.com

  4. compensation claims

    they have began doing this with the main searches on the uk domain for certain keywords. I find it very useful if i haven’t found the resources i need in my initial search.

  5. Prakhar Agarwal

    Okay, it’s nice to see that Step-1 of real image searching seems to be have been accomplished. Now the major work required is to enable user input in the system (that is, Query by Example). We can notice that the input is still “text” and not “image”. Second development would be to set up a relevance feedback system to refine search results with every search. Indeed Google’s main challenge is to use content based retrieval (CBIR) methods and also maintain the scalability. The latest I found during my research was this: http://www-i6.informatik.rwth-aachen.de/~deselaers/fire/
    We will surely see above two developments hitting the web in near future. It’s just about time.
    Cheers,
    Prakhar

Leave a Reply