knowuh



image landscapes:




description: "Image Landscapes" is a shared software application that uses spatial metaphors to help organize and search for digital images. Using models of containment and proximity, users of the system organize their personal photographs by placing images inside each other, or near each other in a pictorial landscape.

The images, and the images' spatial relationships, are stored on a remote database. The server uses the spatial information to infer content relation, and topical similarities of the images. For example, an image of a building may contain images of people who live or work in that building. Images of a mother and a child should be placed near each other, to indicate that the two are related. Once the images have been stored in this way, it becomes much easier to find images based on the content and subject matter.

web start: The application can be launched from your web browser, however there is no documentation at this moment. The webstart version of the application is likely to change without notice. try it anyway!


future directions:


continuum landscape: in the spirit of fuzzy logic, the continuum landscapes provide a way to rank how suitable a given word or set of words is for describing a given image. first the user chooses two antonyms, for instance "hot" and "cold", or "urban" and "rural". the system then finds two images which have been keyed with these descriptions or keywords. A composite landscape is created, which becomes a continuum, in which other images can be placed. In this example, the continuum is a 1 dimensional gradient, but two, or even three dimensional gradients can be imagined. The placement of an image in the continuum field gives the system information about where along the continuum the image belongs, and subsequently, provides the system with additional knowledge about the image. an image can be related in an arbitrary number of landscapes in this way.

time landscape: in the time landscape, the X axis is spatially keyed to automatically harvested (exif) information about when an image was photographed or created. research by microsoft and others has shown that photographs which happen during the same time tend to be 'about' and 'of' the same event, and contain the same subject matters.

narrative landscape: in a narrative landscape, the photographs are used to tell a story. sub regions of the landscape are manually annotated with keywords, and linked to common sense, and word net hierarchies. Images can be placed within this landscape, in various regions. images can be clicked on to get to the next frame of the story.
these narratives can have multiple outcomes, and are non linear. a navigational history (kept at the bottom) records the story as it has progressed so far.

fuzzy ven-searches: after a number of images have been organized in a continuum landscape, the images can then be searched using a boolean fuzzy set search. specify how 'light' or 'happy' or even 'rural' an image you would like in your search results set.

rapid serial visual presentation token in GUI instead of using one image to represent collections of images (spatially organized or temporally) use an rsvp display widget to make navigation easier and faster.