This is a participatory installation piece in which users can draw a continuously growing landscape. The landscape is both geometric and organic--geometric in its use of simple shape primitive and organic in its resemblance of sensed user motion. Aesthetically we intend the tradeoff between these two properties to inform the work's style.

We are interested in the performative aspects of real-time generative motion graphics. In previous projects we have worked with expensive motion capturing systems which are capable of tracking precise positions. One realization from that experience was that precise marker locations is often not necessary for picking up the expressive nature of the performers. A system that specifically looks at the overall motion might be more suitable. This was pretty much the starting point for reapproaching a familiar concept from a different angle.

We started out with an interest in human computer interfaces that are highly intuitive, approachable, and fun to use. The premise is that in many situations it makes sense to trade off precision for simplicity. This also comprises the move to user interfaces that are less intrusive. Sitting down and grabbing a mouse should not be necessary if the application could operate with a wink of a hand. When we started playing around with OpenCV we realized that this can relatively easily be accomplished. We quickly implemented some code that finds the major velocity of motion and used it to rotate a 3D cube. Then we looked at the shapes that can be distilled from the video feed. Both these image analysis techniques are at the basis of the final piece.

Anybody who likes intuitive computer interfaces.

Up to two users enter a 9' x 9' space in front of a display. Their motions get tracked by an organically moving shape on the display. The system has a preference for the users' hands. By holding the rest of the body relatively still the users can start modeling the shape with their hands. From this primary shape the system also extracts the properties for growing and accumulated landscape. Mountain ranges and vegetation flows into the scene.

At the core of the piece is a custom C++ application. It runs on Intel Macs and OSX and requires a Quicktime-compatiple camera. The primary programming libraries are OpenCV and OpenGL. All the motion graphics is implemented with the latter and the computer vision part heavily uses the former. For sensing the user interaction we are creating MHIs (motion history image) and extract vertices from this accumulative motion field.

We are trying to present a human computer interface that stirs up fascination for its extremely intuitive, non-intrusive, and immediate quality. We also hope to awe the audience with the sheer beauty of our abstracted generative landscapes.