I heard about this via a tweet yesterday, view the video here. The video shows off what I think is a pretty amazing concept: First draw a really horrible sketch, then label your illegible doodles with a few descriptive keywords, then some magic turns your sketch into an actual picture. The system apparently works by searching the web or some database of images that will match the keywords you use to label your objects. After images are found they are analyzed by a series of algorithms to help eliminate incorrect matches and to find the matches that were likely intended by the user. These images are then cropped out from their backgrounds using some image manipulation algorithms and the images are laid out on a canvas using the locations that you placed your initial sketches.

What I thought was interesting about this application was the simplicity of the user experience. Anyone can understand it after seeing it once. This seems like it would be a great tool for coming up with simple marketing concepts and brainstorming, or just for the entertainment value.

You can visit the project’s homepage: here

The project page also has links to the source code and to user studies for the project.