Model professor

It was a common meeting of academics. Cindy Grimm, Ph.D., associate professor of computer science and engineering, was wearing protective gloves in the hapkido Korean martial arts class that she helped teach at Brown University, taking the punches of Bill Smart, Ph.D., assistant professor of computer science and engineering.

Grimm was a second-year doctoral student in computer science, Smart a first-year. They barely knew each other. Smart threw an errant punch, causing his elbow to roll off Grimm’s glove and land with a loud, nasty snap on her nose. Smart, flummoxed and sorry that he’d hurt his instructor, offered to drive her to the emergency room.

There they sat, and sat, and sat, and learned a lot about each other. After an exam and X-rays, the intern sent Grimm home. No break. The next morning she got a call from the hospital and was told to come in. The intern had been wrong: She had a broken nose.

Two weeks later, in the same class, Grimm “got Smart,” inadvertently nicking him under the eye and breaking a blood vessel that turned his face into the color of a tornadic sky. Thus, a very strong — and unusual — bond was formed that led to marriage in 2002.

Analyzing real-world data

While Grimm and Smart have set aside the martial arts for the less dynamic sport of rock climbing, they still help each other in the academic ring. Grimm is director of the Department of Computer Science & Engineering’s Media and Machines Laboratory, composed of Smart; Robert Pless, Ph.D., associate professor; Caitlin Kelleher, Ph.D., assistant professor; and Tao Ju, Ph.D., assistant professor. Projects range from computer graphics to computer vision to machine learning, which includes robotics.

“The common thread of our research is that we tend to work with real-world data,” Grimm says. “We take data from sensors, lasers, cameras, CT and MRI machines to present them to other researchers in ways that help them solve their problems. Most of what we try to do is build computer models of the world, based on collected data that we can interact with and explore to better understand what’s going on.”

For example, Grimm is collaborating with Philip Bayly, Ph.D., the Lilyan and E. Lisle Hughes Professor of Mechanical Engineering and chair of the Department of Mechanical, Aerospace & Structural Engineering. Bayly is studying the biomechanics of brain development in the ferret, the smallest mammal that has cortical folds in its brain. These folds are very important in human brain development because some of the most severe neurological problems such as schizophrenia, autism and lissencephaly (a smooth cortex found with severe retardation) are associated with abnormal brain folding. One of Grimm’s students is assembling Bayly’s data to enable a visualization of the mechanical forces at work in folding.

“The long-term goal of this research is to understand how the brain form changes in a particular disease,” Grimm says. “I’m interested in the mathematical question of how to analyze the shapes and track physical points through time automatically. I’ve been mostly pointing Phil and his collaborators toward how the problem looks in graphics.”

Says colleague Pless: “What is special about her is the amount of energy that she pours into research and teaching, supporting colleagues and students. And even more compelling is her ability to translate this energy into lasting contributions.

“She’s now leading an effort to make the entire curriculum more modern, modular and interactive,” Pless says. “As a colleague, she has an intuition that spans from very theoretical to very applied, from formal tools to represent shapes with complex topologies, to insights on what technology demos will resonate with the public.”

Active learning

In 2009, Grimm became principal investigator of a National Science Foundation grant that brings active learning to WUSTL computer science classes, relegating the lecture to being posted on a Web site and viewed the night before in preparation for a class of interaction, group work and peer review. The belief is that active learning, or studio-based instruction, better prepares students for the workplace.

“Active learning is definitely catching on nationally, and the evidence is pretty conclusive that you get lots more engagement in the classroom,” she says. “The syllabus of active learning classes might contain less, but information is retained much better.”

Another area she’s working in is called subtle gaze direction, which is a sneaky way of getting people to draw their gaze on a part of an image without the subject noticing it.

It works like this: If you take an image and make a part of it “blink” — either by making it brighter or a different, gaudy color — your peripheral vision will pick it up and cause your eye to saccade (rapid, intermittent movement) toward that blink. But if the blink is turned off before the eye fully lands on it, the subject doesn’t actually see it.

“The subject’s peripheral vision responds to it, but it doesn’t consciously communicate, ‘There’s a blinky dot, I’m going to look at it,'” Grimm says. “It just gazes over there, so the subject doesn’t actually see it, consciously.”

The potential application for this is as a training mechanism to help technicians screen for tumors.

“People who are really good at this develop certain eye patterns that enable them to look through radiology images to find tumors,” she says. “The thought is you could combine the technique with novel computer vision algorithms to make an estimate of where there might be tumors. There are tons of computer visual algorithms out there, but they’re at most 90 percent accurate. These can give false positives but no false negatives. Ultimately, a human will make the call.”

Art and computers

Grimm grew up in Mountain View, Calif. Her father was a junior-high teacher; her mother worked at home and volunteered as a teacher’s aide at a school that had a high population of disabled children. Her parents took a basic computing class, and her mother began writing software to help the students with grids, maps and simple mathematical graphics.

She took her software to a fledgling company whose people liked her work, and her mother was invited to be a founding member of the educational software company, The Learning Company.

Eventually, her father left teaching and went to work for Apple Computer Inc. In the summers, throughout high school, Grimm tested software for her mother’s company.

This techy background notwithstanding, Grimm entered the University of California, Berkeley, in the fall of 1985 with the intention of becoming an art major. She took a basic computer science class that set the wheels turning: Why not be a double major?

“When I told people that I was majoring in art and computer science, they looked at me as if I’d come from the loony bin,” she says. “So few people were doing that then, but here we routinely turn out several students a year with some combination of art and computer science.”

Art adorns her fifth-floor Lopata Hall office, some of it hers, some her mother’s, who took up drawing later in life. Grimm loves printmaking and figure drawing, both of which figured into a computing challenge she tackled with Smart in 2002.

Smart and his students were building a mobile robot and trying to find a task for it to accomplish. They decided to try to make a robot that could roam a room like a wedding photographer and take pictures. Getting the robot to be mobile was not easy, but neither was teaching it to frame a person in a scene to take a viable picture.

Lewis, the world’s first robotic photographer, made his debut at a cocktail reception at the Clayton Ritz-Carlton hotel in November 2002 at a prestigious meeting of international science writers, the Council for the Advancement of Science Writing’s New Horizons in Science briefing.

The night before, Grimm, Smart and their students went sleepless trying to get Lewis to operate correctly. They transported Lewis (aka “The Big Red Trash Can”) to the Ritz and set up hours before the reception, and as soon as people started arriving at 6 p.m., Lewis took off and roamed the crowd, detecting faces and scenes and computing the right framework. Grimm had written the code that allowed Lewis to become a skilled photographer, based on simple rules of art.

Grimm expects that the shape of things to come for her will be shape understanding.

“We still can’t get quickly from imaging data to geometric models that we can ask interesting questions of,” she says. “That will take time, but there will be some very interesting breakthroughs ahead because there is such an explosion of data available.”