BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030343Z
LOCATION:Meeting Room S422\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251217T163000
DTEND;TZID=Asia/Hong_Kong:20251217T173000
UID:siggraphasia_SIGGRAPH Asia 2025_sess249@linklings.com
SUMMARY:GenAI for Images and Videos
DESCRIPTION:Painting with Paintings: Semantic Stroke-Based Mosaic Generati
 on\n\nWe present a novel mosaic technique that embeds historical artworks 
 within individual brush strokes using semantic segmentation and dual-crite
 ria retrieval to create fluid compositions where each stroke contains auth
 entic artworks.\n\n\nPrudviraj Jeripothula, Abhishek Dangeti, and Vikram J
 amwal (TCS Research)\n---------------------\nBoundary-Conditioned Inpainti
 ng for Constraint-Consistent Procedural Generation Completion\n\nWe presen
 t an inpainting pipeline that creates a failsafe for the wave-function-col
 lapse algorithm by integrating boundary-based diffusion. Our method allows
  for convergence from unsatisfiable constraints caused by incomplete tiles
 ets.\n\n\nZirui Zhang and Kevin Zhang (Purdue University)\n---------------
 ------\nReChar: Revitalising Characters with Structure Preserved and User-
 Specified Aesthetic Enhancements\n\nWe propose ReChar, a framework for art
 istic character generation that explicitly decouples structure, style, and
  decorative elements, enabling controllable generation guided by user-defi
 ned aesthetics.\n\n\nZhongyu Yang (Heriot-Watt University), Junhao Song (I
 mperial College London), Zhang Luo (Seafog AI), Zuhao Yang (Nanyang Techno
 logical University (NTU)), Yang Xu (Southern University of Science and Tec
 hnology), Jingfen Lan (Xidian University), Yonghan Zhang and Wei Pang (Her
 iot-Watt University), Siyang Song (University of Exeter), and Yingfang Yua
 n (Heriot-Watt University)\n---------------------\nChromaFlow: End-to-End 
 Flow Matching for Efficient Video Colorization\n\nWe present ChromaFlow, a
  diffusion-transformer-based video colorization framework that revisits th
 e colorization formulation by using a DCT-based compression mechanism in Y
 CbCr space and directly predicting chrominance components conditioned on l
 uminance.\n\n\nHanyuan Liu and Chengze Li (Saint Francis University), Tien
 -Tsin Wong (Monash University), and Xueting Liu (Saint Francis University)
 \n---------------------\nSkill Transfer Training System for Musculoskeleta
 l Minimally Invasive Treatment Based on Cross-modality Spatial Mapping and
  Integration\n\nWe propose a training platform that synthesizes real-time 
 ultrasound from CT and fuses multimodal imaging to improve spatial underst
 anding, accelerate cross-modality learning, and advance musculoskeletal mi
 nimally invasive surgery skills.\n\n\nZekun Jiang (West China Biomedical B
 ig Data Center, West China Hospital, Sichuan University); Mengqi Tang (Med
 ical Simulation Center of West China Hospital, Sichuan University); Fangwe
 n Huang, Kechen Ren, and Yuhang Liu (West China Hospital, Sichuan Universi
 ty); Kang Li (West China Biomedical Big Data Center, West China Hospital, 
 Sichuan University); and Dan Pu and Xiandi Wang (Medical Simulation Center
  of West China Hospital, Sichuan University)\n\nRegistration Category: Ful
 l Access, Full Access Supporter\n\nSession Chair: Danielle An (Meta)
END:VEVENT
END:VCALENDAR