May 16, 2017
View gallery – 53 images
Art has always been fundamentally intertwined with technology. New techniques and materials have constantly allowed artists to innovate and create new types of works. In this series we look at the impact of digital technologies on art and how artists are creating entirely novel forms of art using these modern tools. We’ve previously examined the fields of ” datamoshing”, ASCII art, BioArt, Minecraft Art and Internet Art. In this instalment we examine a fascinating world where scientists are teaching robots how to paint works of art.
Artificial intelligence systems are currently excelling at producing elaborate digitally generated works of art. Every other week we seem to see a new neural network developed to mimic a famous artists’ aesthetic or convert a photograph into a painterly image. But what about machines actually mimicking the process a human artist uses to paint on a canvas? That particularly human skill seems to be a lot harder for machines to replicate.
NEW ATLAS NEEDS YOUR SUPPORT
It’s just US$19 a year.
In 2016, the RobotArt competition was founded by Stanford educated mechanical engineer Andrew Conru. The competition was designed to stimulate robotic engineers to create new mechanical painting devices. In setting up the competition Conru noted that many of the initial entries were expected to be variations of a simple mechanism where a robotic arm mimics the movements of a human artist, but many teams took the challenge a step further.
The competition saw a variety of different entries, from a team using an eye-tracking system to control a robot’s movement, to a system that had users remotely control a robot via internet-directed brush stroke commands. All the weird and wonderful results reinforced the question of how truly creative a robotically generated work of art could really be.
Below are the recently announced winners of the 2017 RobotArt competition. Be sure to click through to our gallery to get a broader look at each winner’s work.
Winner – PIX18 / Creative Machines lab
From a mechanical engineering team at Colombia University we get the winner of RobotArt 2017, a bot by the name of PIX18. Apparently this is the third generation of a system developed with the goal of creating a robot capable of creating original artwork using the classic medium of oil on canvas.
Judging comments applauded this robot’s ability to produce “some lovely paintings from sources or scratch” and noted that the work had “brush strokes evocative of Van Gogh”.
2nd Place – CMIT ReART
The ReART system uses a haptic recording system to record artists painting a work. The system tracks the position of the brush, the force being exerted and a variety of other data points. A robot then “plays back” the recording, creating a perfectly mimicked ink brush drawing.
The project is from the Department of Electrical Engineering at Kasetsart University in Thailand and looks to develop motion control robotics for a variety of industrial and creative uses.
3rd Place – CloudPainter
CloudPainter is one of the most technically sophisticated projects in the RobotArt competition. Utilizing AI and deep learning systems, the project aims to get the machine to make as many individual creative decisions as possible. According to the creators, currently “the only decision made by a human is the decision to start a painting.”
More info on their process can be found on their website.
One of the judges said of the machine’s work, “Spontaneous paint, “mosaicing” of adjacent tones, layering effects and the graphical interplay between paint strokes of varying textures, are all hand/eye, deeply neurally sophisticated aspects of oil painting…”
4th Place – e-David
e-David is an evolving robotic painting system that uses a visual feedback loop to constantly record and re-process how the machine is interpreting its recreation of an input image. Using an ordinary industrial welding robot combined with cameras, sensors and a control computer, the system can correct errors as it paints, while also understanding what the makers call “human optimization processes”.
5th Place – JACKbDU
This is one of our favorite works from the competition. From a student at New York University Shanghai, this project is inspired by the aesthetic of American artist Chuck Close. The system starts with an input image that is converted to a low resolution and painted pixel by pixel using a mobile robot with omni wheels.
Each oversized, low-res pixel that is cribbled by the robot is roughly the size of a human hand and each entire artwork is 176 X 176 cm (5.7 x 5.7 ft), or just about as tall as a human being.
6th Place – HEARTalion
HEARTalion is a project from Halmstad University in Sweden that attempts to develop a system that can recognize and subsequently depict a person’s emotional state. The system captures emotional signals using a Brain-Machine Interface (BCI) and a robot then attempts to convey the emotions visually based on a model that was developed with advice from two local painters in Halmstead, Peter Wahlbeck and Dan Koon.
One of the impressed RobotArt judges remarked in reference to HEARTalion, “If this body of work was exhibited at a gallery and I was told that the artist aimed to capture emotion through color, composition, and textures – I would buy.”
7th Place – Late Night Projects
This independent entry from an electronic engineer who put in most of the work after his wife and kids had gone to bed uses a simple XYZ axis painter bot guided by two basic behavioral rules. All of this project’s work is from reinterpretations of input images, but because the robot receives no feedback from sensors or cameras, the mixing of colors isn’t faithful to the source. However, the novel strength of this project comes from its gorgeous use of watercolor paint.
8th Place – Wentworth Institute of Technology
Using the precision of a robotic artist to its advantage, this project created a system that minutely controls the pressure and movement of single brush strokes to create stunning images that a human would struggle to accurately produce. The members of the team describe their process in greater detail here and have also publicly offered up their source code in the hope others will build upon their work.
9th Place – CARP
CARP, or Custom Autonomous Robotic Painter, comes from a team at the Worcester Polytechnic Institute in Massachusetts. The system uses image decomposition techniques to dissemble input images, which are then reconstructed by a robot. Visual feedback systems are also incorporated into the process allowing for dynamic corrections to be applied to the work as it is being created.
10th Place – BABOT
An experimental project from a team at MIT. This is an evolving robot arm that was saved from an existence as a decorative coat rack and has slowly been given more peripherals, such as an auto-brush cleaner and wireless control via a video game controller. Equipped with machine learning abilities, the robot can grow its skill set from project to project.
Take a closer look through some more of the amazing and varied robot painted artworks in our gallery.
View gallery – 53 images