The EyeVisionBot at the Centre for Art and Media in Karlsruhe could revolutionise the handling of images: it searches for photo-graphs in databases or on the Internet and is controlled solely by eye movements.
EyeVisionBot
Anyone who needs information from the World Wide Web does not search but instead “googles”. The success of the world’s most popular Internet search engine relies on a sophisticated search-and-find strategy: first Google explores its data libraries for similarities with the required term and sorts the results according to their relevance. A document is regarded as important if it contains the sought term many times, if many links point to it and if it is used frequently. However, this method works well with text documents but hits a boundary where pictures are concerned. As computers have no eyes, they cannot recognise similar visual motifs. Thus, Google can only find images to which text captions were assigned manually.
Dr. Hans Diebner of the Centre for Art and Media (ZKM) in Karlsruhe decided this was too inefficient, and came up with a new strategy to facilitate the search for pictures in a database. Why not let the computer users themselves decide which pictures are important for them – and best do this unconsciously, without tediously clicking around in a database? From this bright idea Diebner developed the EyeVisionBot. It consists of a device to detect eye movements, combined with an algorithm developed by Diebner and his ZKM colleague Lasse Scherffig that uses phrases and image structures to find similar motifs. Visitors to ZKM’s Media Museum, where EyeVisionBot searches through the huge “Media Art Net”, can convince themselves first-hand that the theory does work in practice.
Image selection is surprisingly straightforward, and most importantly works without the standard search templates: to begin with, the monitor shows a selection of arbitrary images. The user glances across them, spending fractions of a second longer on those that are of interest. The system records this fact, and puts together a new selection of images with similar motifs from the database. “After only two or three new selections, the user has bounded the search range to such an extent that the appropriate content has been found”, as Diebner has established in tests.
Currently, EyeVisionBot analyses the time the observer spends looking at images and derives from it the relevance of each. The rule is: the longer an image is observed, the more important it must be. But Diebner is aware that basically we know too little about what goes on in the brain when we perceive pictures: perhaps the interesting ones are those at which observers look for the shortest time, because they recognise them at once as similar to the one they are searching for? Sebastian Fischer from Tübingen’s Knowledge Media Research Centre is investigating this question. Together with Diebner and Scherffig, Fischer received the 2005 doIT Software Award for EyeVisionBot. They are planning a new research project to decipher the brain processes that take place when images are recognised, again using EyeVisionBot – this time not to find pictures but to examine visual perception and the associated effects in the brain.
Diebner is convinced: after centuries dominated by text media such as books and newspapers, this is the dawn of the image-age. The importance of visual media such as television is growing, and with it also the demands on the viewers’ visual skills. No wonder that image science is flourishing – for example in Karlsruhe where besides ZKM, the Graduate School for “anthropology of visual media” at the State University of Arts and Design is also exploring the issue. The combination of art and media technologies at ZKM strikes a chord mainly abroad – “the international resonance is huge”, says ZKM director Peter Weibel. “There is nobody in the art world that doesn’t know and values ZKM”.
Hans Diebner, originally a physicist, wants to continue researching at the interface of perception and technology – in the future at the Institute of New Media in Frankfurt, founded by today’s ZKM director Peter Weibel. Diebner hopes to turn EyeVisionBot into a real product. However, because a vision recording apparatus still costs around 6000 euros, only professional users will be able to enjoy this privilege. Doctors are one possible target group: they analyse endless amounts of x-rays and derive from them clinical findings and the right treatment for their patients. Today, this vast fund of information lies without use in archives and databases. With EyeVisionBot, one could present doctors with x-ray photographs from which they can choose in seconds the ones that most resemble the current case. With those appropriate x-rays, the doctor would be able to find out how other practitioners have dealt with such cases – “a contribution to more objective medical diagnosis”, says Diebner.
But Diebner wants more: one day every PC owner will have an EyeVisionBot, just as today everyone has a mouse. The preparations are well underway: in tests, the team has linked together picture searches in Google with EyeVisionBot. Diebner says: “In the future, everyone will be able to search the Internet for pictures or find archived photographs on the PC – without a single click of the mouse”.
www.medienkunstnetz.de/werke/eye-vision-bot
<- Back to: Success Stories