[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84712-en":3,"doc-seo-84712-105":29,"detail-sidebar-cat-0-en-105":90},{"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":11,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84712,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","Semantic-Guided Progressive Object Removal with Gaussian Splatting","Removing unwanted objects from reconstructed 3D scenes is a key computer-vision problem for AR/VR, robotics, and digital content creation. Existing approaches often fill the entire masked region in a single pass and underuse semantic cues across views, making it hard to preserve complex geometry and textures. This work presents a framework combining Semantic-guided Block Matching (SBM) and Region-Wise Progressive Refinement (RPR) on Gaussian Splatting, improving perceptual quality and cross-view coherence.","Semantic-Guided Progressive Object Removal with Gaussian Splatting  \nXianliang Huang 1 ,3 , Chen Xiao3 , Yuanxiang Ni2 , Guanming Liu3 , Mingkai Liu 1 , Dikai Fan 1 , Xiao Liu 1 , Hao Zhang2 ,†  \narXiv :2607 .04 144v 1 [ cs .RO] 5 Jul 2026  \nAbstract—Removing unwanted objects from reconstructed 3D scenes is an important task in computer vision, supporting applications in AR/VR, robotics, and digital content creation. Existing methods typically complete the entire masked region in a single step and without effectively utilizing semantic information from other views, leading to difficulties in handling complex geometric details and textures. In this work, we propose a novel framework that integrates Semantic-guided Block Matching (SBM) and Region-Wise Progressive Refinement (RPR) for high-quality 3D object removal. First, we leverage DINOv2 to encode semantic guidance from multiview observations, and the best match tokens are decoded to complete missing regions in the target view while maintaining cross-view consistency. Second, we introduce a RPR strategy that segments the target mask into multiple subregions and selectively refines those with poor visual quality. Our method is built upon Gaussian Splatting, ensuring high-fidelity scene reconstruction with efficient computation. Experimental results demonstrate that our approach outperforms existing Gaussianbased methods in terms of perceptual quality and coherence in 3D object removal.  \nI. INTRODUCTION  \nThree-dimensional scene reconstruction and manipulation have been significantly advanced by Neural Radiance Fields [1] and 3D Gaussian Splatting [2] (3DGS), which enable photorealistic and efficient rendering for a wide range of applications, including virtual and augmented reality [3], robotics [4], [5], and autonomous driving [6] . A fundamental yet challenging task in this domain is 3D object removal, which involves eliminating unwanted objects from scenes and realistically completing the resulting holes. This task becomes particularly difficult when removing large objects in unbounded 360° environments, where it is necessary to leverage multi-view observations, hallucinate previously unseen content, and maintain both visual consistency and geometric plausibility across all views. Among recent advances, 3DGS has emerged as a powerful solution for real-time novel view synthesis and editable 3D scene reconstruction. Consequently, accurate and consistent object removal within Gaussian-based representations is becoming increasingly crucial for interactive editing and downstream scene understanding tasks.  \nDespite recent advancements [7], [8], [9], existing 3D object removal methods still face challenges when dealing with complex occlusions and fine-grained geometry. A key limitation lies in their insufficient exploitation of semantic information across multiple views. For instance, methods like  \n1PICO, ByteDance Inc., 2 Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, 3 Fudan University  \n†For correspondence and questions: [h.zhang10@siat.ac.cn](h.zhang10@siat.ac.cn)  \nSPIn-NeRF [7] perform inpainting primarily from 2D inputs while largely neglecting cross-view semantic consistency. As a result, they often produce inconsistent reconstructions, lacking geometric coherence in object regions. Other approaches [10], [11], [12] leverage generative priors through Score Distillation Sampling (SDS) [13] to optimize 3D representations. However, these approaches frequently yield visually inaccurate or overly smooth reconstructions, as they lack explicit geometric guidance and struggle to preserve high-frequency details. Furthermore, these techniques adopt a one-shot completion strategy for the entire masked region, which restricts their ability to iteratively refine suboptimal regions and correct localized artifacts.  \nTo overcome the above limitations, we propose a novel framework that incorporates Semantic-guided Block Matching (SBM) across different views, en","cbCaiqQVakYIaLz8","https://ap.wps.com/l/cbCaiqQVakYIaLz8","pdf",4156625,2,1,"English","en",105,"# Introduction\n## Related Challenges\n## Proposed Framework\n## Semantic-Guided Block Matching (SBM)\n## Region-Wise Progressive Refinement (RPR)\n## Contributions and Experimental Findings","[{\"question\":\"What problem does the proposed method address?\",\"answer\":\"It targets 3D object removal from reconstructed scenes by eliminating unwanted objects and completing the resulting holes with realistic geometry and textures.\"},{\"question\":\"How does Semantic-guided Block Matching (SBM) help?\",\"answer\":\"SBM uses semantic guidance from multiview observations, matching and decoding tokens to complete missing regions in the target view while keeping cross-view consistency.\"},{\"question\":\"What is the role of Region-Wise Progressive Refinement (RPR)?\",\"answer\":\"RPR divides the target mask into subregions and selectively refines those with poor visual quality, focusing computation where fidelity is low to improve realism and coherence.\"}]",1784197791,20,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"semantic-guided-progressive-object-removal-with-gaussian-splatting","",{"@graph":35,"@context":84},[36,52,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,46,49],{"item":40,"name":41,"@type":42,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":20},"https://docshare.wps.com/document/","Document",{"item":47,"name":12,"@type":42,"position":48},"https://docshare.wps.com/document/research-report/",3,{"item":50,"name":13,"@type":42,"position":51},"https://docshare.wps.com/document/semantic-guided-progressive-object-removal-with-gaussian-splatting/84712/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":23,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":40,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does the proposed method address?","Question",{"text":74,"@type":75},"It targets 3D object removal from reconstructed scenes by eliminating unwanted objects and completing the resulting holes with realistic geometry and textures.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does Semantic-guided Block Matching (SBM) help?",{"text":79,"@type":75},"SBM uses semantic guidance from multiview observations, matching and decoding tokens to complete missing regions in the target view while keeping cross-view consistency.",{"name":81,"@type":72,"acceptedAnswer":82},"What is the role of Region-Wise Progressive Refinement (RPR)?",{"text":83,"@type":75},"RPR divides the target mask into subregions and selectively refines those with poor visual quality, focusing computation where fidelity is low to improve realism and coherence.","https://schema.org",{"og:url":50,"og:type":86,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":88,"canonical":50},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,126,129,133],{"id":21,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":28,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":28,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]