Windows on the World: Robert Pless's Research Vision
There are many pictures shared with everyone, they show roads and cities and people and trees and animals. Looking at lots of pictures of places can help show how the world is changing over time, but first you need to know where the picture was taken, what direction it is looking, and what is in the picture. If there are lots and lots and lots and lots of pictures, it is tiring to do this by hand, and we try to use computers to help with this. Sometimes we need more pictures, so we also try to figure out how to get people to help take pictures at the right places when they are walking through the woods or along the water.Webcam Archives and Citizen Science
His biggest project is the Archive of Many Outdoor Scenes (AMOS), which logs images from every publicly available webcam in the world and has now archived over 400 million photos. As he explains: "We log one picture from each camera every half hour; to keep this from being privacy invasive, we avoid cameras that are indoors or can see people from very close. Then, we have projects that use this dataset to measure how the world is changing." For example, there are biology-focused projects like observing when trees leaf out in the spring, when they flower, when they lose their leaves in the fall. The ground-level photos can be compared with satellite estimates. The archive can also be used to explore how people use public space, and Dr. Pless gives one example: "How are walking, biking, and driving habits changing in DC since the start of the bike share program?" AMOS has been saving imagery for 6 years now, so there is often "before" data to go with the "after observations.
Dr. Pless says that he "was amazingly lucky to be at Blair in 1986 when we had a VAX computer and then an SGI graphics workstation in 1989," but computing and imaging technology has progressed at a rapid pace in the intervening decades: "The iPhone I carry completely dominates those systems in computation power. Even more, that computing power is attached to a camera that is better than any digital camera that existed then, and it is connected to the internet to share those images." That progress has made possible many real-world applications for computer vision research: "It is so much fun to work in a field where projects like the Google self-driving car, which were impossible to imagine when it took seconds to read in a 640x480 image, are not only possible but actually really driving around our streets."
When asked about his memories of the Magnet, Dr. Pless recalls the excitement of Mr. Bunday's 9th grade Physics classroom, and he remembers Mr. Donaldson's Optics class as "fantastic." In his own teaching, he has aspired to be like the Magnet teachers who could be "friendly and set really high expectations at the same time." Dr. Pless was also on the tennis team, and he remembers having to take the Metro home after practice and learning a lot of physics on the long ride around the Red Line from Silver Spring to Shady Grove. He is still close friends with his classmate Patrick Baker '90, and they have collaborated on research on persistent surveillance algorithms. Maneesh Agrawala '90 and Alex Berg '93 both work on similar research areas related to Computer Vision, and when asked how the Magnet has helped him in his career, Dr. Pless remarked that "it is amazing how many people I met at Blair that I still interact with professionally."
After earning his Ph.D. from UMD in 2000, he joined the faculty at Washington University in St. Louis and continued to pursue computer science research related to vision. Early in his career, his research was related to human vision: "making Magic Eye pictures (where you stare at the poster and eventually see a 3D dolphin or something) and working on computational explanations of optic illusions." But he moved away from human vision to computer vision, and he hopes that his current work can help to address "bigger problems in the world," like climate change, health and security.
Both of Dr. Pless's parents were academics, and he saw himself on the same career path after he graduated from Cornell and went to the University of Maryland for graduate school. His thesis explored the question: "if your images are corrupted by noise, how does that affect your estimates of the motion in the scene." Dr. Pless explains: "if you want to have a self-driving car, and you put a camera on the front of it, you'd like to know how the car is moving, and what other things in the scene are moving."
After graduating from Blair, Dr. Pless went to Cornell for college. Although he was planning to major in math or physics, a professor sparked his interest in Computer Vision: "My first Computer Science teacher was Dan Huttenlocher ... He was the most charismatic, energetic and interesting teacher I had at Cornell, and that convinced me to think more about Computer Science as a major. His research area was Computer Vision, so when I joined his research lab briefly, that was the first non-class topic I worked on."
Dr. Pless describes his research as "Computer Vision, Remote Sensing and Citizen Science," and he has built tools to allow individuals to create archives of photos over time. He built and shared an app called rePhoto (http://projectrephoto.com), which is available for both iPhone and Android. The rePhoto app makes it easy for anyone to try to repeat a photograph, as he explains: "The app shows the original photograph half transparent so it is easier for you to take the picture from exactly the same viewpoint. Also, there is a database of these projects, so it doesn't have to be the same person that takes the follow on picture, it can be anyone." And there are many groups using this app "to record changes to glaciers, to track the health of urban street trees in New York City, or for more personal uses like documenting house improvements."Silver Spring, Ithaca, College Park, and St. Louis