As Ph.D candidate Susanna Ricco prepares to graduate in summer 2013, she knows she wants to continue working in computer vision or with problems in artificial intelligence.
The idea is to be able to determine and follow the path of any point in a video throughout a sequence. "If you have a video, what I want you to be able to do is to click on any pixel anywhere in that video, then I'll be able to tell you where that point goes throughout the sequence," Ricco said.
The ability to follow the tip of someone's nose or a particular signpost through a video has applications in augmented reality and three-dimensional modeling. Through Ricco's research videos could be edited to change the color of someone's shirt, add writing on a wall or remove an unwanted person or feature. Pixels of a person who disrupts a video sequence by walking through the setting could be replaced with pixels of the background. Ricco's work also would allow a 3-D environment to be constructed using a video's individual frames -- similar to recent commercial photo applications that combine a large number of photos to create a 3-D model of a place, like Microsoft's Photosynth. In addition, Ricco's research could be applied to allow retrieval of videos from YouTube based on the activities they depict.
"Susanna has been working with passion, creativity and dedication on this difficult computer vision problem," Tomasi says. He goes on to describe her work in detail: "Image motion -- also known as optical flow -- is typically computed from a very few frames at a time. In contrast, Susanna takes a long video sequence and stacks up all its frames into a block. Her algorithms then carve this space-time block into three-dimensional 'tubes' one can see when following an image region as it moves through time. In Susanna's world, optical flow is then a fluid of sorts, in which every image point moves together with its neighbors. This fluid is elastic, as it can expand or contract as the camera moves towards or away from an object, or as an object rotates or deforms in the field of view. In addition, the fluid can appear out of nowhere when parts of the background are suddenly revealed from behind an object, or it can disappear instantly, when foreground objects hide previously visible regions of the background. These phenomena make the mathematical description of optical flow a great challenge. Susanna's reformulation of the problem of spatially dense optical flow estimation required intellectual courage back in 2008. She had to rewrite the mathematics of the problem from the ground up, and use computational techniques at the forefront of numerical methods for partial differential equations."
Tomasi praises the excellent mathematical skills Ricco has shown in this work and adds, "She has been rewarded with publications at the top conferences in the field and has become a go-to expert in the area of video analysis. She understands its problems and methods deeply, and she knows the literature intimately."
In addition to her research, Ricco has been passionate about increasing the department's presence at the annual Grace Hopper Celebration of Women in Computing, a conference to increase the visibility of women's contributions. This past fall, after she helped send letters to every female student who had taken a computer science course through the department in the last year, a record 23 students attended the conference with departmental support.
"The goal is to get more women in computer science and retain people already in computer science to try to make the gender balance better than it really is. It's very low," Ricco said. "Since I've been at Duke, the department has become a sponsor for that conference and we send a whole bunch of people every year. I'm really proud of getting the department more involved in that and getting groups of people going every year, because I think it's really important."
Faculty and staff have noted Ricco's support of peers and her help in organizing department endeavors, including serving as the liaison between CS graduate students and faculty. "If you really want something done and you want to know it's going to be done well and on time, you know you can go to her," Graduate Program Coordinator Marilyn Butler said. "She's very involved in the life of the department, both the academic and research aspect as well as the leadership aspect. She's just an overall remarkable person."
As the graduate student liaison a few years ago, Ricco worked extra hard to include graduate student input in key processes like the department's external review and the formulation of a long-term strategic plan. "It was a really cool opportunity to get to be in the faculty meetings, because at that time I was still trying to figure out whether I wanted to go into academia," she recalled.
One of Ricco's favorite activities has included joining about 2,000 other graduate students in K-ville, the annual campus parking-lot campout for a chance to buy basketball season tickets. The self-proclaimed sports fanatic camped out her first semester at Duke in fall 2006. The next year, she took over coordinating the CS grad students' group-try for tickets, only stepping down last fall after five years. "It's a fun way to meet other students and have a 'battle scar' experience," said Ricco, recalling the weekend's 36 hours of virtually no sleep because of the random roll calls that define life in K-ville. Students who achieve the missing of no more than one attendance check are entered into a lottery, and CS students who ultimately walk away with tickets share them with others in the coordinated effort.
As Ricco looks ahead to graduating, she plans to pursue national lab or industry jobs, noting there are a number of interesting problems in computer vision. "And I'd like to do probably all of them," she said.