[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-517650-105":59,"doc-detail-517650-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","meshregen-a-unified-3d-geometry-regeneration-framework","MeshReGen: A Unified 3D Geometry Regeneration Framework","","MeshReGen addresses 3D object regeneration from 2D images and coarse or incomplete initial 3D geometry. Instead of one-shot text/image-to-3D generation with limited control, it conditions a diffusion regenerator on an initial 3D shape, enabling unified tasks such as 3D enhancement, reconstruction, and editing. A VecSet-based conditioning mechanism improves the input geometry with consistent fine-grained details, while self-supervised pretext learning and augmentations train a broadly applicable regeneration prior without extra annotations.",{"@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/meshregen-a-unified-3d-geometry-regeneration-framework/517650/",{"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/meshregen-a-unified-3d-geometry-regeneration-framework/517650.png","ImageObject",300,407,{"name":92,"@type":93},"Valentina","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-19","2026-10-07",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 does MeshReGen generate from and with what conditioning?","Question",{"text":113,"@type":114},"MeshReGen reconstructs complete high-information 3D shapes by conditioning on an initial 3D shape together with a 2D image cue.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How is control achieved compared with one-shot 3D generation?",{"text":118,"@type":114},"Control comes from the conditioning mechanism that updates or improves the provided 3D input geometry, rather than relying only on a text/image prompt.",{"name":120,"@type":111,"acceptedAnswer":121},"How is the 3D regeneration prior trained?",{"text":122,"@type":114},"MeshReGen learns a broadly applicable regeneration prior from off-the-shelf 3D datasets using self-supervised pretext tasks and augmentations, without additional annotations.","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},517650,1791461470,{"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":19,"preview_urls":147},13056703020460,"https://ap-avatar.wpscdn.com/avatar/be000253dac470eee5d?_k=1778207105932848923","MeshReGen: A Unified 3D Geometry Regeneration Framework  \nGeon Yeong Park 1 ,2, Roman Shapovalov2 , Rakesh Ranjan2 , Jong Chul Ye 1,  \nAndrea Vedaldi2, and Thu Nguyen-Phuoc2  \n1 Korea Advanced Institute of Science and Technology (KAIST)  \n2 Meta Reality Labs  \n3D Enhancement  \nFig. 1: MeshReGen is a diffusion-based 3D regeneration framework that reconstructs complete 3D shapes from coarse geometry using 2D image cues. The regeneration prior is learned through self-supervised pretext tasks and augmentations, without taskspecific architectures or extra annotations. Users can input a coarse mesh (3D Enhancement), incomplete point cloud (3D from Point Cloud ), or masked mesh (3D Editing) to recover detailed, complete 3D shapes. Please refer to the project page for more details.  \nAbstract. We consider the problem of regenerating 3D objects from 2D images and initial 3D shapes. Most 3D generators operate in a oneshot fashion, converting text or images to a 3D object with limited controllability. We introduce instead MeshReGen, a 3D regenerator that is conditioned on an initial 3D shape. This conceptually simple formulation allows us to support numerous useful tasks, including 3D enhancement, reconstruction, and editing. MeshReGen uses a new conditioning mechanism based on VecSet, which allows the regenerator to update or improve the input geometry with consistent fine-grained details.  \nMeshReGen learns a widely applicable regeneration prior from off-theshelf 3D datasets via self-supervised pretext tasks and augmentations, without additional annotations. We evaluate both the geometric consistency and fine-grained quality of MeshReGen, achieving state-of-the-art performance in controllable 3D generation across several tasks.  \n2 G. Y. Park et al.  \nFig. 2: Overview. MeshReGen takes both 2D image and initial 3D geometry as input, enabling explicit control over global geometry (e.g., pose, coarse shape) while improving fine-grained details. The 3D condition is encoded as VecSet latents (zL ) that compactly represent the global geometry. After taking the sum with the positional embeddings, these conditioning and random latents are diffused by a DiT into enhanced latents and then decoded into a complete, high-quality 3D shape.  \n1 Introduction  \nWhile the fidelity of 3D generators keeps improving, control is still mostly limited to providing a textual or image prompt, which is often insufficient in applications. For example, 3D artists often start with a 2D concept art that captures the 3D content that must be created, but are often given additional specifications of a different nature too. An example workflow may have the artist sketch the 3D shape using a 3D sculpting tool, followed by iterative refinement or editing. In other cases, the artist starts from a noisy scan (3D point cloud) of a reallife object, obtained using photogrammetry, and reconstructs that into a clean, faithful 3D shape that can be used in production. All these examples require the 3D generation process to be controlled by 3D signals of various types (sculpts, scans, block-outs) in addition to 2D signals (concept art) .  \nRecent advances in 3D generation [56, 64, 76, 79] have focused on an imagebased workflow, where a 3D object xH ⊂ R3 (a surface) is generated from a 2D image I ∈ RD .3 Hence, generation amounts to drawing a sample xH from the distribution p(xH | I) . However, in the more elaborate workflows discussed above, it is necessary to control the generation process by an additional signal yL , like a sculpt, a block-out, or a 3D scan. While the nature and interpretation of these signals are widely different, there are some important similarities too. First, all such 3D signals can be represented as 3D shapes yL ⊂ R3 . Secondly, these  \n3 This is true for text-based generation as well, as the text prompt is converted to a 2D image via a text-to-image model.  \nMeshReGen: A Unified 3D Geometry Regeneration Framework 3  \n3D shapes yL contain significantly less","cbCaikJwYQ9BTIO1","https://ap.wps.com/l/cbCaikJwYQ9BTIO1","pdf",57413985,32,"English","# Overview\n# Introduction\n## Motivation and control limitations\n## Unified 3D shape regeneration formulation\n## Key challenges in training\n## Prior work and gaps\n## VecSet-based latent representation","[{\"question\":\"What does MeshReGen generate from and with what conditioning?\",\"answer\":\"MeshReGen reconstructs complete high-information 3D shapes by conditioning on an initial 3D shape together with a 2D image cue.\"},{\"question\":\"How is control achieved compared with one-shot 3D generation?\",\"answer\":\"Control comes from the conditioning mechanism that updates or improves the provided 3D input geometry, rather than relying only on a text/image prompt.\"},{\"question\":\"How is the 3D regeneration prior trained?\",\"answer\":\"MeshReGen learns a broadly applicable regeneration prior from off-the-shelf 3D datasets using self-supervised pretext tasks and augmentations, without additional annotations.\"}]","MeshReGen: A Unified 3D Geometry Regeneration Framework | PDF",1791343302,81,[148,152,156,160],{"page_no":8,"url":149,"width":150,"height":151},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/95c6ea9e79eb8f06c5a0db4fdbb34834.webp",983,1272,{"page_no":14,"url":153,"width":154,"height":155},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/6d55dfc014a4034ac085b0f60e757268.webp",982,1270,{"page_no":81,"url":157,"width":158,"height":159},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/f5bf8d93dcdc671166cde8a558a0dac0.webp",981,1269,{"page_no":19,"url":161,"width":162,"height":163},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/5129a99fb0dc9299c20afc0d784f1f34.webp",980,1268]