[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84472-en":3,"doc-seo-84472-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84472,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1782698725881665579",8,"Research & Report","Direct Object-Level Reconstruction via Probabilistic Gaussian Splatting","Object-level 3D reconstruction is essential for cultural heritage digitization, industrial manufacturing, and virtual reality, yet existing Gaussian Splatting methods often perform full-scene reconstruction, introducing redundant background and increasing computational and storage overhead. A probabilistic single-object approach based on 2D Gaussian Splatting integrates foreground–background probability cues into Gaussian primitives and prunes low-probability Gaussians during training. YOLO/SAM probability masks enable continuous supervision, dual-stage filtering mitigates background at training startup, and rendered probability masks refine cross-view boundaries. Experiments on MIP-360, T&T, and NVOS show quality comparable to standard 3DGS with about one-tenth Gaussians, demonstrating efficiency and robustness against mask errors.","Direct Object-Level Reconstruction via Probabilistic Gaussian Splatting  \nShuai Guo 1,2 Ao Guo 1,2 Junchao Zhao 1,2 Qi Chen 1,2 Yuxiang Qi3 Zechuan Li 1,2 Dong Chen 1,2 Mingliang Xu 1,2†  \n1 School of Computer and Artificial Intelligence of Zhengzhou University, 450001, China  \n2Zhengzhou University Engineering Research Center of Intelligent Swarm Systems, 450001, China  \n3 School of Data Science and Artificial Intelligence,  \nDongbei University of Finance and Economics, Dalian, 116025, China  \narXiv :2603 . 14316v2 [ cs .CV] 13 Jul 2026  \nAbstract  \nObject-level 3D reconstruction play important roles across domains such as cultural heritage digitization, industrial manufacturing, and virtual reality. However, existing Gaussian Splatting–based approaches generally rely on fullscene reconstruction, in which substantial redundant background information is introduced, leading to increased computational and storage overhead. To address this limitation, we propose an efficient single-object 3D reconstruction method based on 2D Gaussian Splatting. By directly integrating foreground–background probability cues into Gaussian primitives and dynamically pruning lowprobability Gaussians during training, the proposed method fundamentally focuses on an object of interest and improves the memory and computational efficiency. Our pipeline leverages probability masks generated by YOLO and SAM to supervise probabilistic Gaussian attributes, replacing binary masks with continuous probability values to mitigate boundary ambiguity. Additionally, we propose a dual-stage filtering strategy for training’s startup to suppress background Gaussians. And, during training, rendered probability masks are conversely employed to refine supervision and enhance boundary consistency across views. Experiments conducted on the MIP-360, T&T, and NVOS datasets demonstrate that our method exhibits strong self-correction capability in the presence of mask errors and achieves reconstruction quality comparable to standard 3DGS approaches, while requiring only approximately 1/10 of their Gaussian amount. These results validate the efficiency and robustness of our method for single-object reconstruction  \n†Mingliang Xu is the corresponding author. This work was supported in part by the National Natural Science Foundation of China under Grant No. 61903341 and 62325602; in part by the Key Scientific Research Project Plan of Colleges and Universities in Henan Province under Grant No. 26A520039, in part by the Foundation of the State Key Laboratory of Robotics of China under Grant No. 2022-KF-22-06; in part by the National Key Research and Development Program Project under Grant No. 2024YFB3311600 .  \nand highlight its potential for applications requiring both high fidelity and computational efficiency.  \nKeywords: 2D Gaussian Splatting, Single-Object Reconstruction, Probabilistic Supervision, Background Filtering  \n1. Introduction  \nObject-level 3D reconstruction is highly valuable but has not been sufficiently explored in the field of computer vision. The emergence of 3D Gaussian Splatting (3DGS)  \n[17] offers new opportunities to advance research in this domain. 3DGS employs a dense set of Gaussian primitivesto form an explicit scene representation, which improves both training efficiency and reconstruction quality through machine learning. Furthermore, 2D Gaussian Splatting (2DGS) [15] addresses the view-inconsistency problem inherent in 3DGS, representing another major advancement in Gaussian Splatting-based reconstruction methods.  \nHowever, 2DGS, 3DGS, and most subsequent research built upon them primarily focus on full-scene reconstruction, whereas many real-world applications require reconstructing only a specific object. For example, the digital preservation of cultural relics or the geometric modeling of industrial objects typically focuses on specific objects of interest, while regarding the environment as unnecessary background, which incurs significant additio","cbCait15x3KszUhm","https://ap.wps.com/l/cbCait15x3KszUhm","pdf",9997138,2,1,14,"English","en",105,"# Introduction\n## Motivation for single-object reconstruction\n## Background vs object separation challenge\n## Proposed probabilistic 2D Gaussian Splatting framework","[{\"question\":\"Why do full-scene Gaussian Splatting methods cause overhead for single-object tasks?\",\"answer\":\"They include substantial redundant background information, which increases computational cost and storage requirements compared with reconstructing only the object of interest.\"},{\"question\":\"How does the proposed method focus reconstruction on the target object?\",\"answer\":\"It augments each Gaussian primitive with a probabilistic attribute, integrates foreground–background probability cues, and dynamically prunes low-probability Gaussians during training to suppress background modeling.\"},{\"question\":\"How are probabilistic supervision signals generated and used during training?\",\"answer\":\"Probability masks from YOLO and SAM provide pixel-wise likelihoods for belonging to the target object. Continuous probability values replace binary masks to reduce boundary ambiguity, and rendered probability masks are used to refine supervision and improve boundary consistency across views.\"}]",1784195866,35,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"direct-object-level-reconstruction-via-probabilistic-gaussian-splatting","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/direct-object-level-reconstruction-via-probabilistic-gaussian-splatting/84472/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-22","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do full-scene Gaussian Splatting methods cause overhead for single-object tasks?","Question",{"text":75,"@type":76},"They include substantial redundant background information, which increases computational cost and storage requirements compared with reconstructing only the object of interest.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed method focus reconstruction on the target object?",{"text":80,"@type":76},"It augments each Gaussian primitive with a probabilistic attribute, integrates foreground–background probability cues, and dynamically prunes low-probability Gaussians during training to suppress background modeling.",{"name":82,"@type":73,"acceptedAnswer":83},"How are probabilistic supervision signals generated and used during training?",{"text":84,"@type":76},"Probability masks from YOLO and SAM provide pixel-wise likelihoods for belonging to the target object. Continuous probability values replace binary masks to reduce boundary ambiguity, and rendered probability masks are used to refine supervision and improve boundary consistency across views.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]