Ever searched for images based on who or what is in it without help of tags or filenames. it is possible but just as much possible as voice recognition was few years ago.This technology is called Content Based Image Retrieval or CBIR for short.long time ago when my interest in anything related to AI was really big i frequented a number of sites i bumped on a very intersting article Self Organizing Map AI for Pictures By Casey Chesnut .it was reallly good when u consider the accuracy he got with it.He used a type of neural networks called Self Organising Maps (SOM) or Kohonen Map named after the creator of this algorithm.
but since he prefered not to release the source or the program i searched online to find a lot of online sites that perform this kind of search but only online nothing that i can download and use on my large collection of pictures.
Why am i talking about this now ? 
just to say that i have started to work on this and i hope to make a good program and on the way learn a lot of stuff like neural networks and its variants.And i hope to make this my first step into the massive world of AI related technologies . if u are impatient and want to try this now u can try imgseek , a python based program that i tried some time ago. it is good but did not have the accuracy i was expecting and it had some stability problems. but a good software none the less.
siging out for now
Technorati Tags: SOM, kohonen, image+search, , imgseek, CBIR
Well can you give me some details about your approach in CBIR? I am interested and would like to support you. I downloaded imgseek as well but the image import didnt work under windows for dunno which reasons…
[...] . any way it will act as a good continuation along with the other methods of analysis from the previous blog post, using hough Transforms to find edges or more accurately shapes inside images. any more info [...]