Relevant image search algorithm

asked 2015-06-05 06:34:57 -0500


We want to develop a mobile application which helps the user to take a photo of a product (like shirt, belt, shoes etc.) and search the relevant product available in the e-commerce sites. If I took the photo of a shirt with black checks '(front view), it should match the image of similar shirt with black check even if it is side view of that shirt.

We have tried LIRE in Java but its not returning relevant matches we are expecting, some irrelevant products are also listing at the top.

The idea is, there is a search image and a folder of images the search should list the most relevant images.

How can we implement it? Is there any open source algorithm available in Java?

edit retag flag offensive close merge delete


Seriously, google is your best friend. 1 to many matching has been researched 1000x before and there are many possible solutions out there, even based on OpenCV. False matches will always happen, no way to avoid that completely.

StevenPuttemans gravatar imageStevenPuttemans ( 2015-06-05 07:38:12 -0500 )edit