Alumna Receives Best-Paper Award in Vision and Learning

Jan 12, 2015
Catherine Wah

Catherine Wah is a recent CSE alumna (Ph.D. '14) and now a software engineer at Google. She was front and center at the recent Winter Conference on the Applications of Computer Vision (WACV). The 2015 conference took place January 5-9 in Hawaii, where Wah presented a paper on "Learning Localized Perceptual Similarity Metrics for Interactive Categorization" in the session on Vision and Learning. The paper was jointly co-authored with University of Massachusetts-Amherst assistant professor Subhransu Maji and Wah's former doctoral advisor, then-CSE Prof. and current Cornell NYC Tech Prof. Serge Belongie. Wah's research interests are in computer vision, machine learning, and human computation.

According to the abstract for Wah's prize-winning paper, "current similarity-based approaches to interactive finegrained categorization rely on learning metrics from holistic perceptual measurements of similarity between objects or images. However, making a single judgment of similarity at the object level can be a difficult or overwhelming task for the human user to perform. Secondly, a single general metric of similarity may not be able to adequately capture the minute differences that discriminate fine-grained categories. In this work, we propose a novel approach to interactive categorization that leverages multiple perceptual similarity metrics learned from localized and roughly aligned regions across images, reporting state-of-the-art results and outperforming metheods that use a single nonlocalized similarity metric." Wah and her co-authors used examples of bird images compiled by the Caltech-UCSD Visipedia project, including a dataset that Wah herself helped create in hopes of determing a method for computers to recognize different species of bird exclusively through computer vision, or including an element of human crowdsourcing, which could eventually be extrapolated to more robust computer vision systems.

Read the paper "Learning Localized Perceptual Similarity Metrics for Interactive Categorization."