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:20251218T030657Z
LOCATION:Meeting Room S423+S424\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251218T092100
DTEND;TZID=Asia/Hong_Kong:20251218T093200
UID:siggraphasia_SIGGRAPH Asia 2025_sess147_papers_1961@linklings.com
SUMMARY:PractiLight: Practical Light Control Using Foundational Diffusion 
 Models
DESCRIPTION:Yotam Erel (Tel Aviv University), Rishabh Dabral and Vladislav
  Golyanik (Max Planck Institute for Informatics), Amit H. Bermano (Tel Avi
 v University), and Christian Theobalt (Max Planck Institute for Informatic
 s)\n\nLight control in generated images is a difficult task, posing specif
 ic challenges, spanning over the entire image and frequency spectrum. Most
  approaches tackle this problem by training on extensive yet domain-specif
 ic datasets, limiting the inherent generalization and applicability of the
  foundational backbones used. Instead, PractiLight is a practical approach
 , effectively leveraging foundational understanding of recent generative m
 odels for the task. Our key insight is that lighting relationships in an i
 mage are similar in nature to token interaction in self-attention layers, 
 and hence are best represented there. Based on this and other analyses reg
 arding the importance of early diffusion iterations, PractiLight trains a 
 lightweight LoRA regressor to produce the direct irradiance map for a give
 n image, using a small set of training images. We then employ this regress
 or to incorporate the desired lighting into the generation process of anot
 her image using Classifier Guidance. This careful design generalizes well 
 to diverse conditions and image domains. We demonstrate state-of-the-art p
 erformance in terms of quality and control with proven parameter and data 
 efficiency compared to leading works over a wide variety of scenes types. 
 We hope this work affirms that image lighting can feasibly be controlled b
 y tapping into foundational knowledge, enabling practical and general reli
 ghting.\n\nRegistration Category: Full Access, Full Access Supporter\n\nSe
 ssion Chair: Fan Tang (Institute of Computing Technology, Chinese Academy 
 of Sciences)\n\n
END:VEVENT
END:VCALENDAR