Heys @amosbastian thanks for reviewing my work here :)

The work done here is part of a bigger project which i'm working on, specifically for my postgraduate studies..

Personally, i think the biggest challenge i face while implementing this was the idea formulation...
When i first implemented a color detection tool, i realize that it's rather difficult to pin point an exact color name for each color... and especially if you take into consideration of all the color names ..

The tool was too rigid as it only identified a single color given a query ... example: a pink object is somewhat similar to a red object and different people may describe those colors differently...

So when a person describes a pink object, they may actually see it as red... and those objects may not be retrieved in a retrieval system...

In order to make the system somewhat flexible, i decided to compute the similarity score of the query color againt's all other colors... this way, if a pink object was queried, a red object would have high similarity score as well, hence, it would be retrieved!

The following link contains a snippet of how it works in my retrieval system: