BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035307Z
LOCATION:Room 325-AB\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T153000
DTEND;TZID=Asia/Seoul:20221206T170000
UID:siggraphasia_SIGGRAPH Asia 2022_sess157_papers_203@linklings.com
SUMMARY:Text2Light: Zero-shot Text-driven HDR Panorama Generation
DESCRIPTION:Technical Communications, Technical Papers\n\nText2Light: Zero
 -shot Text-driven HDR Panorama Generation\n\nChen, Wang, Liu\n\nHigh-quali
 ty HDRIs (High Dynamic Range Images), typically HDR panoramas, are one of 
 the most popular ways to create photorealistic lighting and 360-degree ref
 lections of 3D scenes in graphics. Given the difficulty of capturing HDRIs
 , a versatile and controllable generative model is highly desired, where l
 ayman users can intuitively control the generation process. However, exist
 ing state-of-the-art methods still struggle to synthesize high-quality pan
 oramas for complex scenes. In this work, we propose a zero-shot text-drive
 n framework, Text2Light, to generate 4K+ resolution HDRIs without paired t
 raining data. Given a free-form text as the description of the scene, we s
 ynthesize the corresponding HDRI with two dedicated steps: 1) text-driven 
 panorama generation in low dynamic range (LDR) and low resolution (LR), an
 d 2) super-resolution inverse tone mapping to scale up the LDR panorama bo
 th in resolution and dynamic range. Specifically, to achieve zero-shot tex
 t-driven panorama generation, we first build dual codebooks as the discret
 e representation for diverse environmental textures. Then, driven by the p
 re-trained Contrastive Language-Image Pre-training (CLIP) model, a text-co
 nditioned global sampler learns to sample holistic semantics from the glob
 al codebook according to the input text. Furthermore, a structure-aware lo
 cal sampler learns to synthesize LDR panoramas patch-by-patch, guided by h
 olistic semantics. To achieve super-resolution inverse tone mapping, we de
 rive a continuous representation of 360-degree imaging from the LDR panora
 ma as a set of structured latent codes anchored to the sphere. This contin
 uous representation enables a versatile module to upscale the resolution a
 nd dynamic range simultaneously. Extensive experiments demonstrate the sup
 erior capability of Text2Light in generating high-quality HDR panoramas. I
 n addition, we show the feasibility of our work in realistic rendering and
  immersive VR.\n\nRegistration Category: FULL ACCESS, ON-DEMAND ACCESS\n\n
 Language: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_203&sess=sess15
 7
END:VEVENT
END:VCALENDAR