Using Machine Learning to Reduce Road Congestion
Student researcher: Malolan Vasu ’23
Major: Computer science and mathematics
Adviser: Christian Lopez, assistant professor of computer science
Malolan Vasu
What my research is about:
“Our project aimed to develop a machine learning model for counting vehicles observed through traffic cameras. Using existing computer vision technology, our model first detected the presence and trajectories of the vehicles in the traffic camera footage. It then determined the path of each vehicle and counted the number of vehicles that followed each path. So far, our research involved testing and evaluating various components of the system to maximize accuracy and reliability. Next, we plan on validating our model on more traffic camera footage and further optimizing our methodology.
“Traffic congestions account for significant losses in time and revenue. With the rising investment in smart cities, we now have a large number of traffic cameras capturing footage. Thus, improving and extending existing computer vision technology can directly aid in traffic management. Within the computer vision field of vehicle counting, most existing models require human annotations of the possible vehicle paths. Our research project aimed to perform the same task without the need for human annotations. This could extend the applicability of such systems to all installed traffic cameras and even rotating traffic cameras. In the future, we plan to incorporate the vehicle counting technology to solve the large umbrella problem of anomaly or accident detection.”
What surprised me the most:
“I often place many academics and researchers on a pedestal: I feel that people like me could never come up with such novel and revolutionary ideas. However, the more time I spent reading about existing research in the field of computer vision, I realized the ideas that seem ‘revolutionary’ to me were incremental changes implemented by researchers whose primary focus was just that; they didn’t just appear out of the blue. I still have tremendous respect for researchers, but I’ve realized that I don’t need to revolutionize a field to meaningfully contribute.”
Why this experience is valuable:
“The research experience taught me how I could better stay motivated and productive when working on projects with such high levels of abstraction. This skill is extremely useful, as people do this all the time in the computer science industry and in academia. Being forced to work on things we don’t understand is quite frustrating, but learning to overcome that frustration is absolutely necessary to gain expertise.”