PhD Candidate
University of Texas at Austin
Areas of Interest
- Artificial Intelligence
Poster
Computational Fashion Understanding
Abstract
Fashion is very much a social and cultural statement. It documents the tastes and values of an era, and even reflects the political orientation and economic development of a nation. Understanding fashion computationally would allow us better personalization, help the industry analyze its enterprise, and bring us closer towards building socially intelligent agents. Unfortunately, current computer vision systems for fashion are lacking in multiple ways. They fail to consider the synergy between multiple garments that compose to form an outfit, clothing's relation to the human wearer, and the underlying social factors that influence our fashion choices. Failing to consider these factors leads to inaccurate, limited, and uninterpretable fashion understanding. The goal of my research is to teach machines to learn fashion notions in the context of outfits, humans, and society. Outfits are about the synergy between multiple garments, and I propose topic-model-based methods that automatically analyze the interplay between garments, and discover underlying combinations that form fashion styles. As outfits are eventually worn by humans, I propose strategies to make fashion recommendation systems aware of an individual’s body shape, and give suggestions on dressing in a more fashionable way. Finally, I explore temporal causality relations between cultural factors and fashion trends to automatically discover influences of society on fashion evolution. All together, I hope my research lays down foundation for human-aware, society-aware, and evolving fashion understanding.
Bio
Wei-Lin Hsiao is a PhD candidate in the Computer Science Department at the University of Texas at Austin, advised by Professor Kristen Grauman. Her research interests are in computer vision, with a focus on computational models for fashion understanding. She develops techniques that recognize fine-grained details in clothing, captures interactions between fashion items and their human wearers, and generalizes to fashion's evolution with society. In addition to her pursuit at UT-Austin, she is also a visiting researcher at Facebook AI Research. She has served as a program committee in conferences including CVPR, ICCV, and ECCV. Her work with collaborators has been covered in Vogue, Elle, and Wired among other fashion magazines.
