E-David. A Painting Process
E-David brings together industrial robot technology, computer graphics, and traditional artistic painting techniques. Instead of virtually simulating each and any kind of images – from finely painted watercolor landscapes to high-end photorealistic game graphics – it incorporates and works with the unpredictability of real material. By opening up the thesaurus of traditional painting techniques, the space of technological image generation is widened, and a completely new class of computer-generated but real paintings is created: you might call them real simulations. To reach this goal, e-David works from both ends at the same time: it adapts computer technology to painting craft and molds artistic handwork to mimic a technical process. One might argue that e-David uses and challenges both, computer graphics and traditional painting.
During painting, e-David is visually supervising itself. This feature separates e-David from the history of painting machines that either work from a strict set of rules – as in baroque human-shaped automata – or trust the machine movements without any further adaptation. See Tinguely’s drawing machines where the hardware construction encodes the painting instruction.
In e-David real media and simulation meet. If one would try to determine every single brush stroke one would end up with completely lifeless paintings. In this sense, every single painting by e-David is unique and original: no other will display the same brushstrokes. In modern and postmodern art, chance plays a decisive role. E-David is a system that is able to balance determination and chance. Moreover, as a hybrid system with a general set-up it can be realized in quite different machines, it is open to a full exploration of the space in between freedom and determination. By that, we might even call it a “philosophical machine” – in the traditional, 19th century meaning of the phrase.
Our new system uses a Kuka Youbot one-arm robot that we use for handling paints. This includes the painting process itself -we apply the brush similar to humans- but also brush cleaning and changing of colours. Since this very small robot is not able to reach a larger area we combine it with a movable canvas. This way we potentially can paint on very large areas using a small robot. Just using an xy-plotter would not allow us to handle paints the same way.
A standard PC controls our robot. After some strokes (~100) have been drawn we take a picture of the canvas. After rectifying the input and colour correction we compare the canvas picture with our given input image. Based on the difference new strokes are computed and drawn in the next iteration.
Inputs to our system are images or images with additional information such as depth values or segmentation information. Having such information we can paint the background first and then realize the foreground, e.g., with smaller brushes.
E-David uses an optimization process that uses visual feedback to realize paintings. This allows us to create paintings with high accuracy and to deal with many practical problems during painting with acrylic paints or other media. Painting styles have to be defined as constraints to this optimization loop. This involves positioning, length and attributes of strokes, colour selection and stroke patterns that can be found in impressionism and expressionism.
The optimization also never results in exactly the same output. Small variations in the used paints, random effects and changes of the optimization parameters lead to different results that, however, all approximate the given input. Once no new strokes can be computed anymore that improve the content of the canvas, the optimization is stopped.
e-David produces paintings with an unique texture and their own charm. Repeated over painting results in very articulated three-dimensional textures on the canvas.
Biography of Prof. Deussen
Deussen graduated from Karlsruhe Institute of Technology in 1996 with a PhD thesis on Computer Graphics. 1996-2000 he was a postdoctoral fellow at Otto-von-Guericke University of Magdeburg, Germany. 2000 he was appointed a professor for Computer Graphics and Media Design at Dresden University of Technology and moved to Konstanz University in 2003, where he is professor for Visual Computing. Coming from technical Computer Graphics, Deussen is interested in creating abstract representations of complex data, which allows to reduce visual information to the essential. Artists have this ability, Deussen tries to learn from them. Being fascinated by painting he decided 2009 to build a painting machine, the e-David Project was born.
Aside from these activities he works on modeling problems in Computer Graphics, on Sampling Theory and on Information Visualization. He has more than 130 publications, Google Scholar lists 5400 citations with an h-index of 35 (by March 2015). He is visiting professor at SIAT (Chinese Academy of Sciences) and won the 1000 talents award in 2014. He is co-Editor in Chief of Computer Graphics Forum, Editor of information Spektrum and speaker of the German Computer Graphics Interest Group within the Gesellschaft für Informatik.