Do you think working with image and video would make an interesting career?
Yuliang Zou definitely does. The first-year Ph.D. student — who would like to join the research arm of a major company one day — is researching computer vision, trying to teach computers to analyze and think like a human when they are given visual data like still images, RGB-D data, or video sequences.
“The computer can recognize objects in the image,” said Zou, who is majoring in computer engineering. “Recent years have witnessed significant progress in this domain as mainstream methodology changes from traditional hand-crafted features to data-driven methods, often referred to as deep learning.
“The main drawback is that we require a lot of annotated data to train the models to perform specific tasks like image classification, object detection, etc. So we are interested in finding an alternative approach to training such models, which can alleviate the requirement of annotations while achieving performance comparable to those models trained with full annotations,” he said.
Last fall, Zou presented “Label-Efficient Learning of Transferable Representations across Domains and Tasks” (collaborating with Stanford University and the University of California, Berkeley) at the 2017 Conference on Neural Information Processing Systems (NIPS) in Long Beach, California, and received a Travel Award from the organization.
Zou’s advisor is Jia-Bin Huang, who was significant in drawing him to Virginia Tech.
“When you are choosing a Ph.D. program, your advisor is the most important factor,” said Zou. “Professor Huang is a rising star in this research area and our research interests are aligned as well.”
This summer, he will intern at Adobe in San Jose, California. Zou anticipates graduating in 2022