Low-Cost Robotic Arm Sketches Faces
ptresset writes "A low-cost robotic arm has been sketching faces at the Kinetica2010 art fair in London. Created by the Aikon project research team, the system drew faces non-stop, its creator having to take the role of an automata to repeatedly change the paper. The Aikon project is based at Goldsmiths College, University of London. The main objective of the Aikon project is to implement a computational system capable of simulating the various important processes involved in face sketching by artists. The ensemble of processes to be simulated include the visual perception the subject and the sketch, the drawing gestures, the cognitive activity, reasoning, the influence of the years of training, etc. It is evident that due to knowledge and technological limitations the implementation of each process will remain coarse and approximate. The system implemented is expected to draw in its own style."
Post.
The robot does fine art but the human changes the paper. Something is wrong here.
Table-ized A.I.
Its not allowed to take photographs in courts here in Victoria, Australia so newspapers employ sketch artists to sit in the court and draw portraits of the accused, witnesses, etc.
(you think I'm kidding, don't you?)
Anyway I wonder if you could take this machine into court and claim that it is only doing what an artist would do.
Incidently some of the artists used recently seem to have been influenced by the impressionist school of drawing because the drawings they make don't always resemble the subject.
http://michaelsmith.id.au
Everybody knows that the artists were starving already, and now they are going to be replaced by robots?!
In all seriousness, though, that is a pretty cool device.
I imagine the algorithims used could have an application to computer facial recognition from video footage.
for a big brother application you would just need a high enough correlation to be passed to a human for final judgement.
...I obey the laws of physics....
There has been too many robot related news here on slashdot. Are the robots taking over the world faster that we expected?
It would be interesting if it could watch you draw, then imitate your style as it draws other subjects. It isn't fun to re-draw the same subject over and over just to see how you progress. Instead you could use it to take each of your drawings and show you, say, a lightbulb would look. Plus it would be fun for people like me who like to draw on occasion but quickly get lazy. That way I could just draw half of something and let the robot finish it. Heck, you could use it to do one of those photo-every-day things, but instead of a photo it's a self-portrait based on your current drawing ability.
Of course, if it became complex enough, it could analyze money and learn how to mimic that drawing style...
My webcomic
Had one back in 1989.
Not that cool really.
The robot arm is straightforward. It looks like it's built out of Dynamixel servos, which are good little programmable servomotors used for the better end of hobbyist robotics. (After 25 years, finally something better than one-way pulse code modulation for hobby servo control. These things use a 1mb/s bidirectional multidrop serial bus.) There are standard brackets for making robot arms and legs, and it looks like they just bolted the thing together from the stock parts kit.
It's not clear how much the software knows about faces. That's the important part. Considerable work has been done on facial feature detection. There are commercial products available. Most of them ignore hairstyle, though, since they're aimed at face recognition.
an automata
I don't think that is the grammatical number you think it is.
Swedish plasma phys. PhD student; MSc EE; knows maths, programming, electronics; finance interest; seeks opportunities
The technical challenge is pretty interesting. I like to see tasks completed by a physical analogue. The art is pretty sketchy, though. I've seen "pencil sketch" photo booths in the mall that make prettier artwork. Those have the advantage of going straight from the converted photograph to a printout of course. When they figure out how to let a machine do shaded sketches with a stick of charcoal, I think that this will be ready to jump from technical oddity to marketable novelty.
Virtue finds and chooses the mean.
Aristotle, Ethica Nichomachea
The 1980's called. They want their pen plotter back.
Kid-proof tablet..
That way we can have self-fabricating robots that also self-assemble themselves.
Sdelat' Ameriku velikoy Snova!
Why does it only draw 2 faces for Blacks ? If Voice of person has Ebonics even a trace
, it always draws same face , looks like Buckwheat
If however they have not a trace of Ebonics in their voice ,
instead, it draws a cookie with a white center and Black outside and spits out the words bad man bad man very bad man Please explain this,
Seriously, no one's posted the Lego version of this already? Actually looks like the Lego bot does better depictions to the (un) trained eye ...
http://robotics.benedettelli.com/portrayer.htm
First came stealing the assembly person's job in auto factories. Now they're stealing the jobs of artists. What's next, a robotic president?
There are some misunderstandings with the article. This is the robotic arm in question. And "Sketch" is Austrian slang for "pulverize."
There is an exhibit in China Science and Technology Museum that also draws pictures. This is made of four ABB robots, so perhaps the cost is a bit more.. =)
Sorry, I could not find any representative pictures what this exhibit draws.
Store with salt
Singular is "automaton", plural is "automata".
I misread the title as "Low-Cost Robotic Arm Stretches Faces." That really didn't sound like it would be very fun.
please look on youtube!!!! http://www.youtube.com/watch?v=AOtQAhblRps
Sorry for the interruption of service http://www.gold.ac.uk/computing/showcaseofwork/aikon-researchprojectpatricktressetandfredericfolleymarie/ http://www.youtube.com/watch?v=AOtQAhblRps
I'd RTFA, but it's down...
What would impress me would be if it did facial recognition to find the face, then tried drawing using the arm, compared the drawing to the image and adjusted as it went until it was finally graded 'acceptable' by a human, at which point it would lock down its algorithm.
I saw this robot at Kinetica. It was not very impressive. First of all, the hardware seemed poorly designed. It was incredibly shaky. When it drew a square around an image, the result were four wobbly lines that did not even connect.
The software, of course, is the more interesting part - and a complete mystery. They had a screen showing a program doing edge detection or something similar on a face. They had a camera that was supposed to capture visitors and draw their portraits. Other than that, nothing was written anywhere about how the software works.
They had some portraits hanging on the walls that looked really nice. But all the ones the robot drew live while I was there came out as a wobbly mess. Maybe it was broken or something... this was on the last day of the exhibition, it could have worn out I guess.
http://74.125.155.132/search?q=cache:bXLtSTi8_nMJ:www.aikon-gold.com/+http://www.aikon-gold.com/&cd=1&hl=en&ct=clnk&gl=us&client=mozilla ...
Ant(Dude) @ Quality Foraged Links (AQFL.net) & The Ant Farm (antfarm.ma.cx / antfarm.home.dhs.org).
I went to Kinetica the other day and saw this and was definitely impressed. What i don't think is obvious from RTFA was that he was running Ubuntu! There were actually 3 individual arms setup, all communicating thru a wireless access point to a laptop he had in front of him. There were even mini postcards he was giving out with a rather dashing interpretation of Alan Turing on it :)
This UID is 7651 digits too high to subjectively infer IQ from.
I guess the real news here is that there are low cost robotic arms.
Looking at the video, I would guess that it's doing some edge detection, stitching edge pixels into chains, and drawing each chain. However, the chains seem to be drawn in arbitrary order, and the robot arm spends most of its time "seeking", pen up, to the start of the next chain. It would have been nice to optimize the drawing order to get faster output. Hell, a human would have drawn the sketch a lot faster.
Of course, obtaining the optimal sequence is equivalent to the traveling salesman problem, but there's lots of cheap and easy-to-implement approximations to the TSP that could have been used.
Alejo