Here you can learn more about the GuitarLab's development and its background which started back in 2006.
The idea for the GuitarLab back in 2006 was very simple:
wouldn't it be cool if when you clicked on a guitar from your favourite artist or a popular guitar from our huge range, you were automatically shown a selection of similar looking guitars?
We were so excited about this idea that we got to work straight away by researching to see if somebody had already had a similar idea and had a solution.
No chance! The only websites we found with a similar idea required you to enter in details manually, which is so difficult when it comes to guitars, as it's so subjective :-)
The first steps
We started by trying out different image recognition algorithms to see if we could try and discover their plus and minus points.
We were however disappointed when we realised that the majority of the algorithms only had disadvantages to offer: The algorithms could recognise colours but due to the fact that guitars are varnished, they will always look different when seen in a different light or angle.
This means that the majority of algorithms can't recognise the same guitar if it is photographed at a slant or in a lighter setting.
In addition to this, a black guitar with a white fretboard is, for every guitarist in the world, black - with nearly every algorithm though, it's white with a black rim.
Also for example with several of the red guitars we tested, we had the problem that at no single point did the guitar resemble a true red, rather it looked white or pink in lighter areas and
nearly black in darker areas.
The shape of a guitar is also an important factor when developing a guitar similiarity search but this wasn't possible with the algorithms.
The analysis of all 5000 guitars and basses we sell would need approximately 9 days and nights on a standard PC.
Why did it have to be guitars?
Image top: Analysis of nearly 5000 colour ranges on one single guitar.
As we didn't have a specialist in this field to help us on the project and there weren't any existing algorithms for us to use, we had to develop our own formular
for the analysis of colour, shape and features of guitars.
In December 2008 we began to put our ideas into practice, but we were always faced with another new hurdle: bad picture material,
blurred photos with a red or green tinge and many compression artifacts. A team of four people worked very hard on this project for several months and in February
2009 they were able to present a beta version of the GuitarLab.
We showed our Beta-Tool version to many guitarists and they were all so impressed by it. It was a win-win situation, we gained ideas from their feedback and they made their
own requests for more available functions on the Guitarlab, such as : Artist Data Base, choice of shape, a front page with quick links to the GuitarLab etc.
It therefore took another month, until we had fulfilled everyone's wishes and the GuitarLab could go live -
nearly three years after the original idea.
We would like to thank the following people:
as without their fantastic support and ideas, it wouldn't have been possible to create the GuitarLab :
- Stefan Gewinner
- Dr. Friedhelm Ernst
- Sebastian Kuhn and Uwe Schlagenhaft from drehmomente.de
- Wolfgang Kraus from Netzmarkt
The GuitarLab Development-Crew: Manuel Ernst, Christian Lager, Sven Schoderböck, Stefan Stammler and Christopher Schirner