Profile

At Sony AI, William is a research scientist in the AI Ethics research flagship. Prior to joining Sony AI, he was a Ph.D. researcher in computer vision at the University of Amsterdam, under the supervision of Cees Snoek (VIS lab). William’s Ph.D. research involved visual search, learning with limited labels, and model biases. Previously, he completed an M.Sc. in biomedical engineering at Polytechnique Montréal, under the supervision of Samuel Kadoury (MedICAL lab) and Chris Pal (Mila). During his graduate studies, William was supported by several scholarships from the Natural Sciences and Engineering Research Council of Canada.

Message

“My work at Sony AI currently focuses on algorithm fairness to enable a better assessment of potential biases and their mitigation in computer vision problems.”

Publications

Content-Diverse Comparisons improve IQA

BMVC, 2022
William Thong, Jose Costa Pereira*, Sarah Parisot*, Ales Leonardis*, Steven McDonagh*

Image quality assessment (IQA) forms a natural and often straightforward undertaking for humans, yet effective automation of the task remains highly challenging. Recent metrics from the deep learning community commonly compare image pairs during training to improve upon trad…

Feature and Label Embedding Spaces Matter in Addressing Image Classifier Bias

BMVC, 2021
William Thong, Cees Snoek*

This paper strives to address image classifier bias, with a focus on both feature and label embedding spaces. Previous works have shown that spurious correlations from protected attributes, such as age, gender, or skin tone, can cause adverse decisions. To balance potential …