Chen Gao traveled to Newcastle, United Kingdom, last month to present a paper on human-object interaction at the 29th British Machine Vision Conference, a major international conference on computer vision and related areas held in the UK.
Gao is a first-year Ph.D. student in the Bradley Department of Electrical and Computer Engineering. After graduating with a master’s degree in electrical and computer engineering from the University of Michigan Ann Arbor in April 2017, Gao came to Virginia Tech as a visiting research assistant to work with Jia-Bin Huang. It was this experience that sparked his interest in the university’s Ph.D. program. Huang, a faculty member at the Discovery Analytics Center, is now his advisor.
“Being a DAC student in the era of big data, I really appreciate that we can benefit from access to massive datasets and develop algorithms to learn patterns,” said Gao, whose career goal is to work in research and development in a company like Facebook or Google.
Gao’s passion for computer vision began with a TED talk by Stanford University Professor Feifei Li. “She spoke about giving sight to machines, teaching them to see and then helping us see better,” he said. “After listening to her, I kept asking myself: ‘How could we teach machines to be more intelligent and, in return, how could computer vision benefit our daily life and the natural environment to create a better future?’”
This question, Gao said, is the motivation behind his “challenging, but very practical” research at DAC. His focus is on human-object interaction detection, a crucial step toward a finer-grained understanding of an image.
“Given an image, we not only detect all the objects in the image, but also detect all the interactions between human and objects. Millions of images are uploaded to social media every day, thus it is essential to cluster images according to the content. Our research provides a potential solution to automatically cluster images according to actions,” he said.
Gao has submitted the paper, Panoramic Robust PCA for Foreground-Background Separation on Noisy, Free-Motion Camera Video, to the IEEE Transactions on Computational Imaging for review. This paper is a journal extension of Augmented Robust PCA for Foreground-Background Separation on Noisy, Moving Camera Video at the 2017 IEEE Global Conference on Signal and Information.
His projected graduation date is June 2022