Student in CSSE earns EPSCoR graduate research scholar award
Published: Jun 1, 2022 8:10 AM
By Joe McAdory
Hai Phan, a fourth-year doctoral student in computer science and software engineering, was one of just 36 graduate students from all academic disciplines in the state to win the Alabama EPSCoR Graduate Research Scholar Round 17 Fellowship – a $25,000 prize.
In collaboration with Anh Nguyen, assistant professor in computer science and software, Phan’s research, “DeepFace-EMD: Re-ranking Using Patch-wise Earth Mover’s Distance Improve Out-Of-Distribution Face Identification,” uses state-of-the-art technology to better distinguish human faces in photographs – particularly in an age of face masks.
“Face identification systems are ubiquitous and they can make costly mistakes when identifying people,” said Phan. “This technology often performs much worse on women, people with a darker skin and faces that are partially occluded with masks or sunglasses."
As a consequence, in 2021, some innocent people were wrongly arrested and denied unemployment benefits as the face identification systems employed by the law enforcement incorrectly identified people.
To improve the accuracy of face identification Phan and Nguyen proposed a novel method to teach Artificial Intelligence systems to learn to compare two faces at multiple stages – comparing two faces as a whole and at fine-grained details.
“For example, our method can ignore unimportant information such as faces with masks or sunglasses and compare important, fine-grained details of a face including the eyes,” Phan said. “Our method is novel and is hypothesized to be more useful to human users than the existing approaches, which do not explicitly enforce multiple-stage comparison.”
Phan is interning for the next six months at Facebook/Meta and Adobe.
“Hai is a talented student with excellent engineering skills, a result-oriented mindset, and always a smile on his face,” said Nguyen, who recently earned a National Science Foundation CAREER Award for his AI/facial recognition research, “Harnessing external knowledge to improve computer vision robustness, explainability, and user accuracy.
“Hai’s research in this EPSCoR proposal will improve the accuracy, usability and efficiency of face identification systems in the wild. I look forward to more of his great work at Auburn and beyond.”
Phan credited his mentors and Auburn University for providing him with an opportunity to pursue his work.
“Auburn Engineering has provided me with the resources and advising I need to succeed,” he said. “This will allow us to prove our research in facial recognition is valid and distinguished from the current facial identification approaches. Thus far, we have found amazing results.”
The Graduate Research Scholars Program, funded through the Alabama Commission on Higher Education, was established in 2006 to:
- Strengthen and enhance the research capacity of Alabama EPSCoR Institutions
- Retain the brightest graduate students for careers in disciplines related to science and engineering in Alabama
- Provide highly trained professionals to support the growth of emerging technology companies in Alabama
According to a note from Alabama EPSCoR, Phan’s award represents a “significant honor and documents that he has now written a successfully funded proposal for external research funding. Listing this award on his resume will provide evidence of his ability to obtain extramural support for research.”
Media Contact: , jem0040@auburn.edu, 334.844.3447
