BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250110T023312Z
LOCATION:Hall B5 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T095800
DTEND;TZID=Asia/Tokyo:20241206T100900
UID:siggraphasia_SIGGRAPH Asia 2024_sess139_papers_273@linklings.com
SUMMARY:Hairmony: Fairness-aware hairstyle classification
DESCRIPTION:Technical Papers\n\nGivi Meishvili, James Clemoes, Charlie Hew
 itt, Zafiirah Hosenie, Xian Xiao, Martin de La Gorce, Tibor Takacs, Tadas 
 Baltrusaitis, Antonio Criminisi, and Chyna McRae (Microsoft); Nina Jablons
 ki (Pennsylvania State University); and Marta Wilczkowiak (Microsoft)\n\nW
 e present a method for prediction of a person's hairstyle from a single im
 age. Despite growing use cases in user digitization and enrollment for vir
 tual experiences, available methods are limited, particularly in the range
  of hairstyles they can capture. Human hair is extremely diverse and lacks
  any universally accepted description or categorization, making this a cha
 llenging task. Most current methods rely on parametric models of hair at a
  strand level.\nThese approaches, while very promising, are not yet able t
 o represent short, frizzy, coily hair and gathered hairstyles. We instead 
 choose a classification approach which can represent the diversity of hair
 styles required for a truly robust and inclusive system. Previous classifi
 cation approaches have been restricted by poorly labeled data that lacks d
 iversity, imposing constraints on the usefulness of any resulting enrollme
 nt system. We use only synthetic data to train our models. This allows for
  explicit control of diversity of hairstyle attributes, hair colors, facia
 l appearance, poses, environments and other parameters. It also produces n
 oise-free ground-truth labels. We introduce a novel hairstyle taxonomy dev
 eloped in collaboration with a diverse group of domain experts which we us
 e to balance our training data, supervise our model, and directly measure 
 fairness. We annotate our synthetic training data and a real evaluation da
 taset using this taxonomy and release both to enable comparison of future 
 hairstyle prediction approaches. We employ an architecture based on a pre-
 trained feature extraction network in order to improve generalization of o
 ur method to real data and predict taxonomy attributes as an auxiliary tas
 k to improve accuracy. Results show our method to be significantly more ro
 bust for challenging hairstyles than recent parametric approaches. Evaluat
 ion with taxonomy-based metrics also demonstrates the fairness of our meth
 od across diverse hairstyles.\n\nRegistration Category: Full Access, Full 
 Access Supporter\n\nLanguage Format: English Language\n\nSession Chair: Ku
 i Wu (LightSpeed Studios)
URL:https://asia.siggraph.org/2024/program/?id=papers_273&sess=sess139
END:VEVENT
END:VCALENDAR