[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-328054-105":59,"doc-detail-328054-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","rendering-aware-bayesian-3d-gaussian-splatting-with-native-uncertainty-and-adaptive-complexity-control","Rendering-Aware Bayesian 3D Gaussian Splatting with Native Uncertainty and Adaptive Complexity - Control","","3D Gaussian splatting (3DGS) delivers real-time, high-fidelity novel-view synthesis, but standard training outputs point estimates and hand-tuned heuristics without native uncertainty or principled complexity control. A rendering-aware Bayesian framework is introduced using a Normal-Inverse-Wishart (NIW) posterior over Gaussian means and covariances, with optional Dirichlet-process complexity signaling. Posterior re-rendering yields calibrated predictive uncertainty to support active view selection under fixed budgets. Experiments show PSNR and LPIPS improvements versus acquisition baselines, with substantially reduced 95% coverage error at lower training cost.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/rendering-aware-bayesian-3d-gaussian-splatting-with-native-uncertainty-and-adaptive-complexity-control/328054/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/rendering-aware-bayesian-3d-gaussian-splatting-with-native-uncertainty-and-adaptive-complexity-control/328054.png","ImageObject",300,407,{"name":92,"@type":93},"Quinn","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-09","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",14,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does the paper address in standard 3D Gaussian splatting training?","Question",{"text":113,"@type":114},"Standard 3DGS relies on point estimates and hand-tuned heuristics, lacking native uncertainty characterization and offering no principled way to control or evaluate complexity or weakly supported geometry under limited views or fixed budgets.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does the proposed Bayesian method model uncertainty and complexity in 3DGS?",{"text":118,"@type":114},"It embeds the 3DGS renderer in a probabilistic observation model, placing a Normal-Inverse-Wishart (NIW) posterior over each Gaussian’s mean and covariance. An optional Dirichlet-process extension turns component usage into a probabilistic complexity signal, enabling rendering-aware complexity control.",{"name":120,"@type":111,"acceptedAnswer":121},"How is predictive uncertainty used for active view selection and what are the reported benefits?",{"text":122,"@type":114},"Re-rendering posterior geometry samples produces calibrated predictive uncertainty that selects informative views near high-variance regions. Under a fixed-budget 16-to-32 task, the method improves PSNR and reduces LPIPS relative to standard ensemble baselines, while also reducing 95% coverage error compared to proxy and deep-ensemble approaches.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},328054,1790814768,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":145,"read_time":146,"preview_page_count":4},962075114765,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","Rendering-Aware Bayesian 3D Gaussian Splatting with Native Uncertainty and Adaptive Complexity  \nControl  \nGaoxiang Jia∗  \nAdvanced Micro Devices, Inc. [jiagaoxiang@gmail.com](jiagaoxiang@gmail.com)  \nVikram Appia  \nAdvanced Micro Devices, Inc.  \narXiv :2607 .05522v 1 [ cs .CV] 6 Jul 2026  \nJunzhou Huang  \nDept. of Computer Science and Engineering University of Texas at Arlington  \nXinlei Wang∗  \nDept. of Mathematics & Division of Data Science University of Texas at Arlington [xinlei.wang@uta.edu](xinlei.wang@uta.edu)  \nAbstract  \n3D Gaussian splatting (3DGS) is a strong representation for real-time, high-fidelity novel-view synthesis, yet its standard training pipeline relies on point estimates and hand-tuned heuristics that do not natively provide uncertainty or principled complexity control. This limitation is most apparent under limited view support or fixed acquisition budgets, where a point estimate cannot identify weakly supported geometry or rank informative candidate views. We introduce a rendering-aware Bayesian framework that tracks 3DGS geometry with a Normal-Inverse-Wishart (NIW) posterior over Gaussian means and covariances using renderer-derived surrogate summaries; an optional Dirichlet-process extension turns component usage into a probabilistic complexity signal, and the training schedule makes the closed-form-versus-approximate inference boundary explicit. Re-rendering posterior geometry samples produces native predictive uncertainty for interval calibration and active view selection. In a fixed-budget 16-to-32 active-view-selection task, native NIW acquisition improves PSNR by +0 .453 dB and LPIPS by −0 .0146 over a scoring-only 3-member standard-ensemble acquisition baseline, winning 29/39 scene-seed pairs and 10/13 scene means on PSNR, with PSNR gains also over PPU-style (+0 .355 dB) and NIW proxy (+0 .401 dB) acquisition. NIW native intervals reduce 95% coverage error by about 17× relative to a shared proxy (|Cov@95 − 0 .95| : 0.046 versus 0.796) and are about 10× closer to nominal coverage than a 3-member deep ensemble (0 .047 versus 0.454) at roughly one-third the training cost. As a reconstruction compatibility check, paired NIW-vs-standard analysis over 39 scene-seed runs yields +0 .030 dB PSNR (95% cluster bootstrap CI [+0 .004 , +0 .059] dB) for 1.6% additional training time. Together, these results position Bayesian 3DGS as a practical probabilistic scene representation for decision-facing tasks such as active view selection.  \n1 Introduction  \n3D Gaussian splatting (3DGS) has become a central representation for novel-view synthesis, combining strong reconstruction quality with real-time rendering (Kerbl et al., 2023) . Yet the standard  \n∗Co-corresponding authors.  \nPreprint.  \na Standard 3DGS  \n{μk, Σk } fixed  \npoint-estimate Gaussians  \nd Method overview  \nb Bayesian NIW  \nposterior  \nq(μk, Σk) = NIW(⋅)   \nover geometry  \nN calibrated views  \nRGB images + camera poses  \nLikelihood  \nNIW  \nPixel UQ  \nIj, p ∼􎈺(Cj, p , σ2I3 )  \nalpha-composited Gaussian renderer  \nq(μk, Σk) = NIW(⋅) up = Vars [p(s)]  \nconjugate update from bright = high pixel surrogate stats posterior uncertainty  \nRenderer-derived surrogates → NIW geometry posterior → native pixel uncertainty  \nStandard uniform  \nStandard ensemble Standard PPUNIW uniform  \nNIW proxy  \nNIW native (ours)  \ne Final PSNR after acquisition ( ↑ )  \n19.8 20.0 20.2 20.4 20.6 20.8 21.0 21.2  \nFigure 1: Rendering-aware Bayesian 3DGS. (a) Standard 3DGS keeps {µk , Σk } as point estimates.(b) Bayesian 3DGS-NIW attaches a posterior q(µk , Σk); translucent clouds show posterior geometry draws, with tighter clouds for better-supported components. (c) In fixed-budget 16→32 active-viewselection, native uncertainty selects views near high-variance regions. (d) Camera-calibrated views define a rendering likelihood, renderer-derived surrogate statistics feed the NIW update, and posterior re-rendering yields pixel uncertainty. (e) Final PSNR reaches 20.84 dB, +0 .45","cbCairefsMqTMptr","https://ap.wps.com/l/cbCairefsMqTMptr","pdf",533727,26,"English","# Abstract\n# 1 Introduction\n## Rendering-aware Bayesian 3DGS overview\n## NIW posterior and surrogate summaries\n## Predictive uncertainty for active view selection\n## Experimental results and reconstruction checks","[{\"question\":\"What problem does the paper address in standard 3D Gaussian splatting training?\",\"answer\":\"Standard 3DGS relies on point estimates and hand-tuned heuristics, lacking native uncertainty characterization and offering no principled way to control or evaluate complexity or weakly supported geometry under limited views or fixed budgets.\"},{\"question\":\"How does the proposed Bayesian method model uncertainty and complexity in 3DGS?\",\"answer\":\"It embeds the 3DGS renderer in a probabilistic observation model, placing a Normal-Inverse-Wishart (NIW) posterior over each Gaussian’s mean and covariance. An optional Dirichlet-process extension turns component usage into a probabilistic complexity signal, enabling rendering-aware complexity control.\"},{\"question\":\"How is predictive uncertainty used for active view selection and what are the reported benefits?\",\"answer\":\"Re-rendering posterior geometry samples produces calibrated predictive uncertainty that selects informative views near high-variance regions. Under a fixed-budget 16-to-32 task, the method improves PSNR and reduces LPIPS relative to standard ensemble baselines, while also reducing 95% coverage error compared to proxy and deep-ensemble approaches.\"}]","Rendering-Aware Bayesian 3D Gaussian Splatting with Native Uncertainty and Adaptive Complexity - Control | PDF",1789978579,66]