Slashdot Mirror
Robot Produces Paintings With That 'Imperfect' Human Look
kkleiner writes "An artistic robotic system named e-David has been developed that produces paintings that appear to be created by humans. Using an iterative process of brush strokes and image comparison, e-David's assembly line welder arm can paint in up to 24 colors and add shading where needed. The robot even cleans its five brushes along the way, according to University of Konstanz researchers who developed the system as an exercise in machine learning."
how it models imperfection so perfectly
Actually it's more efficient. By allowing the analog medium to introduce imperfections you don't have to.
A fool throws a stone into a well and a thousand sages can not remove it.
This lacks one vital component: Creativity
A painter may think "I may want to make that woman's eyes a bit more smiling", and then do so. Or think "If I add a stone fence between the buildings, it will look more severe".
Or even "the sky would look better with a green streak".
So while this might be a nice exercise in machine learning, don't insult its good workmanship by calling it art.
I don't mean to take away form the robotics work or the research, but the headline appears to be jumping the gun. Most of the sample paintings look like GIMP filters or that machine at Chuck-E-Cheese that draws the kids' pictures while they wait.
I was expecting a flexible arm mimicking Monet's technique or something. At this point I'd be much happier with an elephant painting on my wall - it's more "human" than the robot's.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
I've got to say, I'm surprised at the negative comments thus far. A lot of engineering creativity obviously went into creating this robot, do you people not appreciate that? Do you have any idea how hard it is to get something this elaborate to work reliably? Have any of you naysayers actually tried to build even a simple robot?
You call it a simple plotter. Really? A plotter is basically a dumb printer. This is is not a dumb printer. It creates the image iteratively by examining what it looks like and modifying as appropriate.
BTW, the tree's shadow would be roughly consistent with the sun being about 15 degrees above the left horizon.
soylentnews.org
Interesting, but it's still not that much different from a printer with an algorithm to imitate a painterly look. There is software (like Corel Painter) that can transform photos to look like they were painted using different mediums. I could load a photo, use an automated feature in Painter, and print it, and it would basically do the same thing as this robot.
Robot has a hobby
Builds Volkswagens by daylight
Paints people at night
That robot will be rejected from a Fine Arts academy, failing the entrance exam twice, and from disgust and despise will try to take over the world instead.
"BTW, the tree's shadow would be roughly consistent with the sun being about 15 degrees above the left horizon."
You are in serious need of a guide dog.
This concept is nothing new - As a musician, I know that the majority of digital recording software supports adjustable rhythm "randomization" to "humanize" drumming, for example. You start with machine-perfect rhythm, then slide the knob down a tad to "Neil Peart," then near the bottom of the knob's travel you get "sloppy drunk John Bonham."
-- Ethanol-fueled
The irony is that while the robot has been perfecting the human look my wife has perfected that machine look. Her drawings look so real that people mistake them for photographs. She does edit out and add in but it comes out looking so real it is mistaken for reality.
As a someone with a Masters of Fine Art in painting, I can tell you there is not a lot of interest relating to art.
First: "Our hypothesis is that painting ... can be seen as optimization processes in which color is manually distributed on a canvas until the painter is able to recognize the content" is off base
All the lines in all the work are all the same length and thickness. Almost no artist simple distributes color. Artist chose details and focus.In this case David is being helped because it is using composed photography to copy.
Second: Even if they could get close to copying human style, it is not that interesting precisely because it is following an algorithm. The idea "the machine might enable new techniques since labor plays no role any more" is pretty weak. Artists typical employ computers to do what a computer does well, not to imitate humans. It is quite possible someone will actually do precisely what the authors suggest and use the machines ability for work without rest. There are always artist who find ways to use tools in new ways or to use them to make commentary on the process. This puts the robot in the same league as a chainsaw for carving wood, or paint that drips down from a rope.
As someone who as worked with machine learning a bit, there is not a huge amount of interest here either.
All in all it was probably fun and interesting to work on, but not all the interesting to read about or watch.
Copying stroke for stroke is a different thing altogether. There is a whole industry for this. http://www.artsstudio.com/ Price ranges with quality. Genuine paintings done by hand go from $200 to somewhere around $10,000 to $15,000 I think. They are not priceless. There is something about human nature the values the original. The price of art is a pure economic ideal. It is worth exactly what someone is willing to pay for it, so you can't really argue that someone overpaid.
The high end copies entail using the same techniques and materials which can be quite laborious. Some material are hand made and recreation requires a lot of specialized knowledge practice. Working with the material also takes lots of skill and practice. Glazing techniques, etc take a long time are more that stroke copy. Even if the robot can make the exact marks, the materials will come from someone else,
So if the robot is very good a stroke for stroke copy it would be better than what the low end people are producing. However, making the material and some techniques are probably outside a stroke for stroke copy. So I estimate the value at $500.