Researchers at the Max Planck Institute have developed a method that makes it possible to sort large sets of pictures on-screen very rapidly. Their algorithm could well have some very useful applications in the e-commerce field.
Visual recognition is something that human beings are rather good at. We are for instance usually able to sort into the same category pictures of animals of the same species or objects of the same type. However, this everyday skill is much harder for computers. While our brains register and recognise a very large number of factors and pointers almost immediately, formalising these criteria into computer software is very difficult. Or at least is has been so far.
Now however researchers at the Max Planck Institute for Informatics in Saarbrücken, Germany appear to have gone a long way towards solving this problem. The research team have developed software that enables you to sort a lot of visual images on the screen at high speed. If this seems a fairly trivial benefit, it could nevertheless prove very effective in helping to enhance the visual architecture of e-commerce sites.
Choose your own sorting criteria
What the German researchers have done in order to get around the traditional obstacles to machine visual recognition is to integrate a plane-by-plane recognition capability into the algorithmic analysis. The programme they have developed allows them to categorise the data according to pre-defined characteristics. Now users of the programme will be able to choose from a wide range of criteria offered by the software and give them a higher or lower weighting so as to create their own classification, i.e. their own fully personalised visual arrangement of the items. Explains Max Planck team member Tobias Ritschel: “All the user has to do is to click on three or more pictures with the mouse and drag them to certain points on the screen… The programme then calculates the requested order and arranges the pictures correspondingly.”
Creating a harmonious display
The software not only represents a major advance from a scientific and technical viewpoint, but may well also have significant implications for online product display. The algorithm arranges the images according to certain visual features such as size and intensity and then distributes them consistently over the free space on the screen in such a way that they create a harmonious overall picture. The programme could therefore provide a number of options for e-commerce. Rather than a mesh structure, which is very formal and also very inflexible, the Max Planck team’s software will be able to enhance the effectiveness of the images on screen in terms of consumer impact by providing more simultaneous visual information. Bernhard Reinert, who co-developed the software programme, highlights the flexibility which the algorithm offers: “On the one hand, e-tailers will be able to represent their products in a more compact and visually pleasing way than before. On the other hand, our programme works very flexibly, so that web designers can easily create their websites for different end-user devices such as computers or smartphones”.