Category Archives: Computer Vision

CHIMP Wins Third Place at DARPA Robotics Challenge

CHIMP, a four-limbed robot designed and built by Carnegie Mellon University’s Tartan Rescue Team, finished third and won $500,000 June 6 at the DARPA Robotics Challenge (DRC), a two-day event that pitted 24 of the world’s most advanced robots against each other in a test of their ability to respond to disasters.

During its best run, the robot engaged the enthusiastic audience as it overcame several mishaps and missteps to complete all eight of the possible tasks in 55 minutes, 15 seconds — good enough to put the team in first place on the first day of competition.

CHIMP was the only robot of the many that fell during the competition that was able to get back on its feet unassisted.

“Our team operating the robot kept their cool,” said Tony Stentz, team leader and research professor. “They managed to get CHIMP to recover and complete all of the tasks. It says a lot about the robot and a lot about the people. It means there’s great promise for this technology.”

 

Breathe Cam Provides People With Tool To Study The Air We Breathe

A system of four cameras, called Breathe Cam, now keeps a constant watch on air quality over Pittsburgh, providing citizens with a new interactive tool for monitoring and documenting visual pollution in the air they breathe and even tracing it back to its sources.

Funded by The Heinz Endowments as part of its Breathe Project, the camera system was developed and deployed by the CREATE Lab in Carnegie Mellon University’s Robotics Institute. Anyone can access the cameras online at http://breatheproject.org/learn/breathe-cam, where images of the Downtown, East End and Mon Valley skylines are updated around the clock.

Using the interactive controls, people can zoom in on items of interest, whether it’s a hovering brown cloud or individual smokestacks or coke plants. They can scan back in time to observe changes in visibility or to try to find the sources of dirty air. They also can skip back to particular dates and times that have been catalogued since the cameras were installed.

The researchers also have developed a computer vision tool to help people identify and quantify events of interest, such as releases from a smokestack. Users can correlate the visual conditions with hourly reports of fine particulate matter, ozone and other pollutant levels recorded by Allegheny County Health Department air monitoring stations.

Popular Science Honors Four Projects as “Best of What’s New”

Four inventions that trace their origins to the School of Computer Science and, particularly, the Robotics Institute, have been honored by the annual Best of What’s New Awards by Popular Science.

This year’s winners, published in the magazine’s December issue, include the Flex System, a neck surgery tool based on snake robot research; 360fly, a panoramic video camera; 3-D Object Manipulation Software, a photo editing tool, and LiveLight, a method for automatically editing out the boring parts of personal or security videos.

Martial Hebert Named Robotics Institute Director

Martial Hebert, a leading researcher in computer vision and robotics at Carnegie Mellon University since 1984, will become director of the university’s Robotics Institute.

“Martial is a widely admired and respected leader in robotics,”said Andrew Moore, dean of the School of Computer Science. “Over the years, he and the people who have worked with him have produced some of the most impactful work on robot vision and sensing that the world has seen. We are all very excited to have him lead one of CMU’s most important centers of excellence.”

A native of Chatou, France, who earned a doctorate in computer science at the University of Paris, Hebert joined the Robotics Institute in 1984, just five years after the Robotics Institute was founded, and was named a full professor in 1999. The Robotics Institute has since grown into the world’s largest robotics education and research institution, with an annual research budget of more than $54 million.

“Having joined the institute shortly after its creation, I am honored and thrilled to now have the opportunity to work with the faculty, students and staff to shape the next phase of its journey,” Hebert said.

Robotics Institute Computer Teaches Itself Common Sense

A computer program called the Never Ending Image Learner (NEIL) is running 24 hours a day at Carnegie Mellon University, searching the Web for images, doing its best to understand them on its own and, as it builds a growing visual database, gathering common sense on a massive scale.

NEIL leverages recent advances in computer vision that enable computer programs to identify and label objects in images, to characterize scenes and to recognize attributes, such as colors, lighting and materials, all with a minimum of human supervision. In turn, the data it generates will further enhance the ability of computers to understand the visual world.

But NEIL also makes associations between these things to obtain common sense information that people just seem to know without ever saying — that cars often are found on roads, that buildings tend to be vertical and that ducks look sort of like geese. Based on text references, it might seem that the color associated with sheep is black, but people — and NEIL — nevertheless know that sheep typically are white.

“Images are the best way to learn visual properties,” said Abhinav Gupta, assistant research professor in Carnegie Mellon’s Robotics Institute. “Images also include a lot of common sense information about the world. People learn this by themselves and, with NEIL, we hope that computers will do so as well.”

WorldKit: Ad Hoc Interactive Applications on Everyday Surfaces

Creating interfaces in the world, where and when we need them, has been a persistent goal of research areas such as ubiquitous computing, augmented reality, and mobile computing. The WorldKit system makes use of a paired depth camera and projector to make ordinary surfaces instantly interactive. Using this system, touch-based interactivity can, without prior calibration, be placed on nearly any unmodified surface literally with a wave of the hand, as can other new forms of sensed interaction. From a user perspective, such interfaces are easy enough to instantiate that they could, if desired, be recreated or modified “each time we sat down” by “painting” them next to us. From the programmer’s perspective, our system encapsulates these capabilities in a simple set of abstractions that make the creation of interfaces quick and easy. Further, it is extensible to new, custom interactors in a way that closely mimics conventional 2D graphical user interfaces, hiding much of the complexity of working in this new domain.