{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T09:42:21Z","timestamp":1777110141722,"version":"3.51.4"},"reference-count":37,"publisher":"Wiley","license":[{"start":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T00:00:00Z","timestamp":1777075200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T00:00:00Z","timestamp":1777075200000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Computer Graphics Forum"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>The abstraction of 3D objects with simple geometric primitives like cuboids allows us to infer structural information from complex geometry. It is important for 3D shape understanding, structural analysis and geometric modeling.<\/jats:p>\n                  <jats:p>We introduce a novel fine\u2010to\u2010coarse self\u2010supervised learning approach to abstract collections of 3D shapes. Our architectural design allows us to reduce the number of primitives from hundreds (fine reconstruction) to only a few (coarse abstraction) during training. This allows our network to optimize the reconstruction error and adhere to a user\u2010specified number of primitives per shape while simultaneously learning a consistent structure across the whole collection of data. We achieve this through our abstraction loss formulation which increasingly penalizes redundant primitives. Furthermore, we introduce a reconstruction loss formulation to account not only for surface approximation but also volume preservation. Combining both contributions allows us to represent 3D shapes more precisely with fewer cuboid primitives than previous work.<\/jats:p>\n                  <jats:p>We evaluate our method on collections of man\u2010made and humanoid shapes comparing with previous state\u2010of\u2010the\u2010art learning methods on commonly used benchmarks. Our results confirm an improvement over previous cuboid\u2010based shape abstraction techniques. Furthermore, we demonstrate our cuboid abstraction in downstream tasks like clustering, retrieval, and partial symmetry detection.<\/jats:p>","DOI":"10.1111\/cgf.70344","type":"journal-article","created":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T08:56:23Z","timestamp":1777107383000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Self\u2010supervised Learning of Fine\u2010to\u2010Coarse Cuboid Shape Abstraction"],"prefix":"10.1111","author":[{"ORCID":"https:\/\/orcid.org\/0009-0008-0672-7565","authenticated-orcid":false,"given":"Gregor","family":"Kobsik","sequence":"first","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0006-1567-4667","authenticated-orcid":false,"given":"Morten","family":"Henkel","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0451-671X","authenticated-orcid":false,"given":"Yanjiang","family":"He","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-0918-7998","authenticated-orcid":false,"given":"Victor","family":"Czech","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1309-8003","authenticated-orcid":false,"given":"Tim","family":"Elsner","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8160-2315","authenticated-orcid":false,"given":"Isaak","family":"Lim","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7880-9470","authenticated-orcid":false,"given":"Leif","family":"Kobbelt","sequence":"additional","affiliation":[{"name":"Visual Computing Institute RWTH Aachen University  Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"311","published-online":{"date-parts":[[2026,4,25]]},"reference":[{"key":"e_1_2_7_2_2","doi-asserted-by":"crossref","DOI":"10.1037\/0033-295X.94.2.115","article-title":"Recognition-by-components: a theory of human image understanding","volume":"94","author":"Biederman I.","year":"1987","journal-title":"Psychological review"},{"key":"e_1_2_7_3_2","volume-title":"Sensor fusion IV: control paradigms and data structures","author":"Besl P. J.","year":"1992"},{"key":"e_1_2_7_4_2","doi-asserted-by":"crossref","unstructured":"BogoF. RomeroJ. Pons-MollG. BlackM. J.: Dynamic FAUST: Registering human bodies in motion. InIEEE Conf. on Computer Vision and Pattern Recognition (CVPR)(2017). 5","DOI":"10.1109\/CVPR.2017.591"},{"key":"e_1_2_7_5_2","doi-asserted-by":"crossref","unstructured":"ChenZ. ChenQ. ZhouH. ZhangH.: Dae-net: Deforming auto-encoder for fine-grained shape co-segmentation. InACM SIGGRAPH 2024 Conference Papers(2024). 3 6","DOI":"10.1145\/3641519.3657528"},{"key":"e_1_2_7_6_2","volume-title":"ShapeNet: An Information-Rich 3D Model Repository","author":"Chang A. X.","year":"2015"},{"key":"e_1_2_7_7_2","unstructured":"ChevalierL. JailletF. BaskurtA.:Segmentation and superquadric modeling of 3d objects. 2"},{"key":"e_1_2_7_8_2","doi-asserted-by":"crossref","unstructured":"ChenZ. TagliasacchiA. ZhangH.: Bsp-net: Generating compact meshes via binary space partitioning. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2020). 7","DOI":"10.1109\/CVPR42600.2020.00012"},{"key":"e_1_2_7_9_2","unstructured":"ChenZ. YinK. FisherM. ChaudhuriS. ZhangH.: Bae-net: Branched autoencoder for shape co-segmentation. InProceedings of the IEEE\/CVF International Conference on Computer Vision(2019). 2 3 6"},{"key":"e_1_2_7_10_2","unstructured":"DosovitskiyA.: An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929(2020). 3"},{"key":"e_1_2_7_11_2","unstructured":"FerreiraP.: Sampling superquadric point clouds with normals.arXiv preprint arXiv:1802.05176(2018). 3"},{"key":"e_1_2_7_12_2","unstructured":"FedeleE. SunB. GuibasL. PollefeysM. EngelmannF.: SuperDec: 3D Scene Decomposition with Superquadric Primitives. InProceedings of the IEEE\/CVF International Conference on Computer Vision (ICCV)(2025). 3 6"},{"issue":"6","key":"e_1_2_7_13_2","article-title":"Sdm-net: Deep generative network for structured de-formable mesh","volume":"38","author":"Gao L.","year":"2019","journal-title":"ACM Transactions on Graphics (TOG)"},{"key":"e_1_2_7_14_2","doi-asserted-by":"crossref","unstructured":"LienJ.-M. AmatoN. M.: Approximate convex decomposition. InProceedings of the twentieth annual symposium on Computational geometry(2004). 2","DOI":"10.1145\/997817.997889"},{"issue":"6","key":"e_1_2_7_15_2","article-title":"Seg-mat: 3d shape segmentation using medial axis transform","volume":"28","author":"Lin C.","year":"2020","journal-title":"IEEE transactions on visualization and computer graphics"},{"key":"e_1_2_7_16_2","unstructured":"LiS. PaschalidouD. GuibasL.: Pasta: Controllable part-aware shape generation with autoregressive transformers.arXiv preprint arXiv:2407.13677(2024). 2"},{"key":"e_1_2_7_17_2","doi-asserted-by":"crossref","unstructured":"LiuW. WuY. RuanS. ChirikjianG. S.: Robust and accurate superquadric recovery: A probabilistic approach. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2022). 2 6","DOI":"10.1109\/CVPR52688.2022.00270"},{"key":"e_1_2_7_18_2","doi-asserted-by":"crossref","unstructured":"LiJ. WangH. TanJ. YuanJ.: Shared latent membership enables joint shape abstraction and segmentation with deformable superquadrics.IEEE Transactions on Image Processing(2024). 3 4","DOI":"10.1109\/TIP.2024.3404234"},{"issue":"4","key":"e_1_2_7_19_2","article-title":"Grass: Generative recursive autoencoders for shape structures","volume":"36","author":"Li J.","year":"2017","journal-title":"ACM Transactions on Graphics (TOG)"},{"issue":"6","key":"e_1_2_7_20_2","article-title":"Structurenet: hierarchical graph networks for 3d shape generation","volume":"38","author":"Mo K.","year":"2019","journal-title":"ACM Transactions on Graphics (TOG)"},{"key":"e_1_2_7_21_2","doi-asserted-by":"crossref","unstructured":"MitraN. J. WandM. ZhangH. Cohen-OrD. KimV. HuangQ.-X.: Structure-aware shape processing. InACM SIGGRAPH 2014 Courses.2014. 2","DOI":"10.1145\/2614028.2615401"},{"key":"e_1_2_7_22_2","doi-asserted-by":"crossref","unstructured":"NiuC. LiM. XuK. ZhangH.: Rim-net: Recursive implicit fields for unsupervised learning of hierarchical shape structures. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2022). 3 6","DOI":"10.1109\/CVPR52688.2022.01148"},{"key":"e_1_2_7_23_2","unstructured":"PiluM. FisherR. B.:Equal-distance sampling of superellipse models. 3"},{"key":"e_1_2_7_24_2","doi-asserted-by":"crossref","unstructured":"PaschalidouD. GoolL. V. GeigerA.: Learning unsupervised hierarchical part decomposition of 3d objects from a single rgb image. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2020). 3","DOI":"10.1109\/CVPR42600.2020.00114"},{"key":"e_1_2_7_25_2","doi-asserted-by":"crossref","unstructured":"PaschalidouD. KatharopoulosA. GeigerA. FidlerS.: Neural parts: Learning expressive 3d shape abstractions with invertible neural networks. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2021). 3","DOI":"10.1109\/CVPR46437.2021.00322"},{"key":"e_1_2_7_26_2","doi-asserted-by":"crossref","unstructured":"PaschalidouD. UlusoyA. O. GeigerA.: Superquadrics revisited: Learning 3d shape parsing beyond cuboids. InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition(2019). 2 3 4 6","DOI":"10.1109\/CVPR.2019.01059"},{"key":"e_1_2_7_27_2","article-title":"Pointnext: Revisiting pointnet++ with improved training and scaling strategies","volume":"35","author":"Qian G.","year":"2022","journal-title":"Advances in neural information processing systems"},{"key":"e_1_2_7_28_2","unstructured":"QiC. R. SuH. MoK. GuibasL. J.: Pointnet: Deep learning on point sets for 3d classification and segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition(2017). 2"},{"key":"e_1_2_7_29_2","doi-asserted-by":"crossref","DOI":"10.1111\/cgf.13323","volume":"37","author":"Rodrigues R. S.","year":"2018","journal-title":"Computer Graphics Forum"},{"issue":"2","key":"e_1_2_7_30_2","doi-asserted-by":"crossref","DOI":"10.1109\/34.44401","article-title":"Recovery of parametric models from range images: The case for superquadrics with global deformations","volume":"12","author":"Solina F.","year":"1990","journal-title":"IEEE transactions on pattern analysis and machine intelligence"},{"issue":"6","key":"e_1_2_7_31_2","article-title":"Learning adaptive hierarchical cuboid abstractions of 3d shape collections","volume":"38","author":"Sun C.-Y.","year":"2019","journal-title":"ACM Transactions on Graphics (TOG)"},{"key":"e_1_2_7_32_2","unstructured":"ShuaiQ. ZhangC. YangK. ChenX.: Dpf-net: Combining explicit shape priors in deformable primitive field for unsuper-vised structural reconstruction of 3d objects. InProceedings of the IEEE\/CVF International Conference on Computer Vision(2023). 3 5 6 7 11 12"},{"key":"e_1_2_7_33_2","doi-asserted-by":"crossref","unstructured":"TulsianiS. SuH. GuibasL. J. EfrosA. A. MalikJ.: Learning shape abstractions by assembling volumetric primitives. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition(2017). 2 3","DOI":"10.1109\/CVPR.2017.160"},{"key":"e_1_2_7_34_2","doi-asserted-by":"crossref","unstructured":"WuR. ZhuangY. XuK. ZhangH. ChenB.: Pq-net: A generative part seq2seq network for 3d shapes. InProceedings of the IEEE\/CVF conference on computer vision and pattern recognition(2020). 2","DOI":"10.1109\/CVPR42600.2020.00091"},{"issue":"4","key":"e_1_2_7_35_2","doi-asserted-by":"crossref","DOI":"10.1145\/3450626.3459873","article-title":"Unsupervised learning for cuboid shape abstraction via joint segmentation from point clouds","volume":"40","author":"Yang K.","year":"2021","journal-title":"ACM Transactions On Graphics (TOG)"},{"key":"e_1_2_7_36_2","doi-asserted-by":"crossref","unstructured":"YeJ. HeY. ZhouY. ZhuY. XiaoK. LiuY.-J. YangW. HanX.:Primitiveanything: Human-crafted 3d primitive assembly generation with auto-regressive transformer 2025. 2","DOI":"10.1145\/3721238.3730732"},{"key":"e_1_2_7_37_2","doi-asserted-by":"crossref","unstructured":"YiL. KimV. G. CeylanD. ShenI.-C. YanM. SuH. LuC. HuangQ. ShefferA. GuibasL.: A scalable active framework for region annotation in 3d shape collections.ACM Transactions on Graphics (ToG)(2016). 2","DOI":"10.1145\/2980179.2980238"},{"key":"e_1_2_7_38_2","article-title":"3dilg: Irregular latent grids for 3d generative modeling","volume":"35","author":"Zhang B.","year":"2022","journal-title":"Advances in Neural Information Processing Systems"}],"container-title":["Computer Graphics Forum"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70344","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1111\/cgf.70344","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70344","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T08:56:27Z","timestamp":1777107387000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/cgf.70344"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,25]]},"references-count":37,"alternative-id":["10.1111\/cgf.70344"],"URL":"https:\/\/doi.org\/10.1111\/cgf.70344","archive":["Portico"],"relation":{},"ISSN":["0167-7055","1467-8659"],"issn-type":[{"value":"0167-7055","type":"print"},{"value":"1467-8659","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,25]]},"assertion":[{"value":"2026-04-25","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70344"}}