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:Auditorium\, Level 5\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T100000
DTEND;TZID=Asia/Seoul:20221206T120000
UID:siggraphasia_SIGGRAPH Asia 2022_sess153_papers_203@linklings.com
SUMMARY:Text2Light: Zero-shot Text-driven HDR Panorama Generation
DESCRIPTION:Technical Papers\n\nText2Light: Zero-shot Text-driven HDR Pano
 rama Generation\n\nChen, Wang, Liu\n\nHigh-quality HDRIs (High Dynamic Ran
 ge Images), typically HDR panoramas, are one of the most popular ways to c
 reate photorealistic lighting and 360-degree reflections of 3D scenes in g
 raphics. Given the difficulty of capturing HDRIs, a versatile and controll
 able generative model is highly desired, where layman users can intuitivel
 y control the generation process. However, existing state-of-the-art metho
 ds still struggle to synthesize high-quality panoramas for complex scenes.
  In this work, we propose a zero-shot text-driven framework, Text2Light, t
 o generate 4K+ resolution HDRIs without paired training data. Given a free
 -form text as the description of the scene, we synthesize the correspondin
 g HDRI with two dedicated steps: 1) text-driven panorama generation in low
  dynamic range (LDR) and low resolution (LR), and 2) super-resolution inve
 rse tone mapping to scale up the LDR panorama both in resolution and dynam
 ic range. Specifically, to achieve zero-shot text-driven panorama generati
 on, we first build dual codebooks as the discrete representation for diver
 se environmental textures. Then, driven by the pre-trained Contrastive Lan
 guage-Image Pre-training (CLIP) model, a text-conditioned global sampler l
 earns to sample holistic semantics from the global codebook according to t
 he input text. Furthermore, a structure-aware local sampler learns to synt
 hesize LDR panoramas patch-by-patch, guided by holistic semantics. To achi
 eve super-resolution inverse tone mapping, we derive a continuous represen
 tation of 360-degree imaging from the LDR panorama as a set of structured 
 latent codes anchored to the sphere. This continuous representation enable
 s a versatile module to upscale the resolution and dynamic range simultane
 ously. Extensive experiments demonstrate the superior capability of Text2L
 ight in generating high-quality HDR panoramas. In addition, we show the fe
 asibility of our work in realistic rendering and immersive VR.\n\nRegistra
 tion Category: FULL ACCESS, EXPERIENCE PLUS ACCESS, EXPERIENCE ACCESS, TRA
 DE EXHIBITOR\n\nLanguage: ENGLISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_203&sess=sess15
 3
END:VEVENT
END:VCALENDAR