[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-189127-105":53,"doc-detail-189127-en":126},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":119,"head_meta":121,"extra_data":123,"updated_unix":125},105,"en","neurips-2022-get3d-a-generative-model-of-high-quality-3d-textured-shapes-learned-from-images-paper-conference","NeurIPS 2022 Get3D - A Generative Model of High-Quality 3D Textured Shapes Learned from Images - Paper Conference","","Quantitative comparisons evaluate 3D generative methods that differ in representation, supervision, and surface properties, including implicit 3D, point clouds, and 2D neural fields paired with textured meshing. The results report coverage (COV), distribution distances (MMD with LFD/CD), and generation quality (FID) across multiple categories such as Car, Chair, Mbike, and Animal, using metrics for both original and 3D evaluations. The proposed approach achieves consistently strong coverage and lower MMD/FID, with further improvements from variants like Ours+Subdiv. and Ours (improved G).",{"@graph":63,"@context":118},[64,80,101],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":51,"@type":70,"position":76},"https://docshare.wps.com/template/general/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/neurips-2022-get3d-a-generative-model-of-high-quality-3d-textured-shapes-learned-from-images-paper-conference/189127/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/neurips-2022-get3d-a-generative-model-of-high-quality-3d-textured-shapes-learned-from-images-paper-conference/189127.png","ImageObject",442,249,{"name":88,"@type":89},"Hazel","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-09-27","2026-09-03",true,{"@type":98,"interactionType":99,"userInteractionCount":79},"InteractionCounter",{"@type":100},"ViewAction",{"@type":102,"mainEntity":103},"FAQPage",[104,110,114],{"name":105,"@type":106,"acceptedAnswer":107},"How does Get3D generate high-quality 3D textured shapes, and how is it represented?","Question",{"text":108,"@type":109},"Get3D is evaluated as a 3D generation approach using mesh-based representation and textured outputs. The comparison includes methods using implicit 3D representations, point clouds, or neural fields, highlighting differences in textured mesh and topology handling.","Answer",{"name":111,"@type":106,"acceptedAnswer":112},"Which evaluation metrics are used to compare methods?",{"text":113,"@type":109},"Methods are compared using COV (coverage), MMD (with LFD and CD components), and FID (reported for both original and 3D settings). These metrics are shown for multiple categories to assess geometric diversity and visual quality.",{"name":115,"@type":106,"acceptedAnswer":116},"Does the proposed method outperform baselines across object categories?",{"text":117,"@type":109},"Across Car, Chair, Mbike, and Animal, the proposed method (“Ours”) shows strong COV and generally lower MMD/FID than several baselines. Variants such as Ours+Subdiv. and Ours (improved G) further improve some metrics.","https://schema.org",{"og:url":78,"og:type":120,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":122,"canonical":78},"index,follow",{"doc_id":124,"site_id":56},189127,1789974689,{"code":4,"msg":5,"data":127},{"doc_id":124,"user_id":128,"nickname":88,"user_avatar":129,"doc_module":9,"category_id":50,"category_name":51,"doc_title":59,"doc_description":61,"doc_content":130,"file_id":131,"file_url":132,"file_type":133,"file_size":134,"view_count":79,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":20,"language":135,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":136,"faqs":137,"seo_title":138,"seo_description":61,"update_tm":139,"read_time":47},137441390410,"https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984","| Method | Application | Representation | Supervision | Textured mesh | Arbitrary topology |\n| --- | --- | --- | --- | --- | --- |\n| OccNet [38] | 3D generation | Implicit | 3D | ✗ | ✓ |\n| PointFlow [59] | 3D generation | Point cloud | 3D | ✗ | ✓ |\n| Texture3D [48] | 3D generation | Mesh | 2D | ✓ | ✗ |\n| StyleNerf [23] | 3D-aware NV | Neural field | 2D | ✗ | ✓ |\n| EG3D [8] | 3D-aware NV | Neural field | 2D | ✗ | ✓ |\n| PiGAN [7] | 3D-aware NV | Neural field | 2D | ✗ | ✓ |\n| GRAF [52] | 3D-aware NV | Neural field | 2D | ✗ | ✓ |\n| Ours | 3D generation | Mesh | 2D | ✓ | ✓ |\n\n| Category | Method | COV (%, ↑) |  | MMD (↓) |  | FID (↓) |  |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  | LFD | CD | LFD | CD | Ori | 3D |\n| Car | PointFlow [59] | 51.91 | 57.16 | 1971 | 0.82 | - | - |\n|  | OccNet [38] | 27.29 | 42.63 | 1717 | 0.61 | - | - |\n|  | Pi-GAN [7] | 0.82 | 0.55 | 6626 | 25.54 | 52.82 | 104.29 |\n|  | GRAF [52] | 1.57 | 1.57 | 6012 | 10.63 | 49.95 | 52.85 |\n|  | EG3D [8] | 60.16 | 49.52 | 1527 | 0.72 | 15.52 | 21.89 |\n|  | Ours | 66.78 | 58.39 | 1491 | 0.71 | 10.25 | 10.25 |\n|  | Ours+Subdiv. | 62.48 | 55.93 | 1553 | 0.72 | 12.14 | 12.14 |\n|  | Ours (improved G) | 59.00 | 47.95 | 1473 | 0.81 | 10.60 | 10.60 |\n| Chair | PointFlow [59] | 49.58 | 71.87 | 3755 | 3.03 | - | - |\n|  | OccNet [38] | 61.10 | 67.13 | 3494 | 3.98 | - | - |\n|  | Pi-GAN [7] | 53.76 | 39.65 | 4092 | 6.65 | 65.70 | 120.53 |\n|  | GRAF [52] | 50.23 | 39.28 | 4055 | 6.80 | 43.82 | 61.63 |\n|  | EG3D [8] | 58.31 | 50.14 | 3444 | 4.72 | 38.87 | 46.06 |\n|  | Ours | 69.08 | 69.91 | 3167 | 3.72 | 23.28 | 23.28 |\n|  | Ours+Subdiv. | 71.59 | 70.84 | 3163 | 3.95 | 23.17 | 23.17 |\n|  | Ours (improved G) | 71.96 | 71.96 | 3125 | 3.96 | 22.41 | 22.41 |\n\n\n| Category | Method | COV (%, ↑) |  | MMD (↓) |  | FID (↓) |  |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  | LFD | CD | LFD | CD | Ori | 3D |\n| Mbike | PointFlow [59] | 50.68 | 63.01 | 4023 | 1.38 | - | - |\n|  | OccNet [38] | 30.14 | 47.95 | 4551 | 2.04 | - | - |\n|  | Pi-GAN [7] | 2.74 | 6.85 | 8864 | 21.08 | 72.67 | 131.38 |\n|  | GRAF [52] | 43.84 | 50.68 | 4528 | 2.40 | 83.20 | 113.39 |\n|  | EG3D [8] | 38.36 | 34.25 | 4199 | 2.21 | 66.38 | 89.97 |\n|  | Ours | 67.12 | 67.12 | 3631 | 1.72 | 65.60 | 65.60 |\n|  | Ours+Subdiv. | 63.01 | 61.64 | 3440 | 1.79 | 54.12 | 54.12 |\n|  | Ours (improved G) | 69.86 | 65.75 | 3393 | 1.79 | 48.90 | 48.90 |\n| Animal | PointFlow [59] | 42.70 | 74.16 | 4885 | 1.68 | - | - |\n|  | OccNet [38] | 56.18 | 75.28 | 4418 | 2.39 | - | - |\n|  | Pi-GAN [7] | 31.46 | 30.34 | 6084 | 8.37 | 36.26 | 150.86 |\n|  | GRAF [52] | 60.67 | 61.80 | 5083 | 4.81 | 42.07 | 52.48 |\n|  | EG3D [8] | 74.16 | 58.43 | 4889 | 3.42 | 40.03 | 83.47 |\n|  | Ours | 79.77 | 78.65 | 3798 | 2.02 | 28.33 | 28.33 |\n|  | Ours+Subdiv. | 66.29 | 74.16 | 3864 | 2.03 | 28.49 | 28.49 |\n|  | Ours (improved G) | 74.16 | 82.02 | 3767 | 1.97 | 27.18 | 27.18 |\n\n| Class | Img Res | COV (%, ↑) |  | MMD (↓) |  | FID (↓) |\n| --- | --- | --- | --- | --- | --- | --- |\n|  |  | LFD | CD | LFD | CD |  |\n| Car | 1282 | 9.28 | 8.25 | 2224 | 1.30 | 39.21 |\n|  | 5122 | 52.32 | 44.13 | 1593 | 0.80 | 13.19 |\n|  | 10242 | 66.78 | 58.39 | 1491 | 0.71 | 10.25 |\n| Chair | 1282 | 38.25 | 33.98 | 3886 | 5.90 | 43.04 |\n|  | 5122 | 68.80 | 69.92 | 3149 | 3.90 | 30.16 |\n|  | 10242 | 69.08 | 67.87 | 3167 | 3.74 | 23.28 |\n| Mbike | 5122 | 68.49 | 65.75 | 3421 | 1.74 | 74.04 |\n|  | 10242 | 67.12 | 64.38 | 3631 | 1.73 | 65.60 |\n| Animal | 5122 | 77.53 | 78.65 | 3828 | 2.01 | 29.75 |\n|  | 10242 | 79.78 | 78.65 | 3798 | 2.03 | 28.33 |","cbCaitDrF8LA22ka","https://ap.wps.com/l/cbCaitDrF8LA22ka","pdf",13985157,"English","# Method Comparisons\n## Representations and Supervision\n## Metrics and Category Results\n# Performance Across Categories\n## Car\n## Chair\n## Mbike\n## Animal","[{\"question\":\"How does Get3D generate high-quality 3D textured shapes, and how is it represented?\",\"answer\":\"Get3D is evaluated as a 3D generation approach using mesh-based representation and textured outputs. The comparison includes methods using implicit 3D representations, point clouds, or neural fields, highlighting differences in textured mesh and topology handling.\"},{\"question\":\"Which evaluation metrics are used to compare methods?\",\"answer\":\"Methods are compared using COV (coverage), MMD (with LFD and CD components), and FID (reported for both original and 3D settings). These metrics are shown for multiple categories to assess geometric diversity and visual quality.\"},{\"question\":\"Does the proposed method outperform baselines across object categories?\",\"answer\":\"Across Car, Chair, Mbike, and Animal, the proposed method (“Ours”) shows strong COV and generally lower MMD/FID than several baselines. Variants such as Ours+Subdiv. and Ours (improved G) further improve some metrics.\"}]","NeurIPS 2022 Get3D - A Generative Model of High-Quality 3D Textured Shapes Learned from Images - Paper Conference | PDF",1788394482]