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:20241205T164100 DTEND;TZID=Asia/Tokyo:20241205T165300 UID:siggraphasia_SIGGRAPH Asia 2024_sess136_papers_306@linklings.com SUMMARY:DIScene: Object Decoupling and Interaction Modeling for Complex Sc ene Generation DESCRIPTION:Technical Papers\n\nXiao-Lei Li (BNRist, Department of Compute r Science and Technology, Tsinghua University); Haodong Li (Hong Kong Univ ersity of Science and Technology, Guangzhou); and Hao-Xiang Chen, Tai-Jian g Mu, and Shi-Min Hu (BNRist, Department of Computer Science and Technolog y, Tsinghua University)\n\nThis paper reconsiders how to distill knowledge from pretrained 2D diffusion models to guide 3D asset generation, in part icular to generate complex 3D scenes: it should accept varied inputs, i.e. , texts or images, to allow for flexible expression of requirement; object s in the scene should be style-consistent and decoupled with clearly model ed interactions, benefiting downstream tasks. \n We propose DIScene, a nov el method for this task.It represents the entire 3D scene with a learnable structured scene graph: each node explicitly models an object with its ap pearance, textual description, transformation, geometry as a mesh attached with surface-aligned Gaussians; the graph's edges model object interacti ons. \n With this new representation, objects are optimized in the canonic al space and interactions between objects are optimized by object-aware re ndering to avoid wrong back-propagation. \n Extensive experiments demonstr ate the significant utility and superiority of our approach and that DISc ene can greatly facilitate 3D content creation tasks.\n\nRegistration Cate gory: Full Access, Full Access Supporter\n\nLanguage Format: English Langu age\n\nSession Chair: Manolis Savva (Simon Fraser University) URL:https://asia.siggraph.org/2024/program/?id=papers_306&sess=sess136 END:VEVENT END:VCALENDAR