{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T18:50:38Z","timestamp":1775501438395,"version":"3.50.1"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1014047","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T00:00:00Z","timestamp":1775433600000}}],"reference-count":41,"publisher":"Public Library of Science (PLoS)","issue":"3","license":[{"start":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T00:00:00Z","timestamp":1773360000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000001","name":"National Science Foundation","doi-asserted-by":"publisher","award":["BCS-2142269"],"award-info":[{"award-number":["BCS-2142269"]}],"id":[{"id":"10.13039\/100000001","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Both humans and deep learning models can recognize objects from 3D shapes depicted with sparse visual information, such as a set of points randomly sampled from the surfaces of 3D objects (termed a point cloud). Although deep learning models achieve human-like performance in recognizing objects from 3D shapes, it remains unclear whether these models develop 3D shape representations similar to those used by human vision for object recognition. Evidence suggests that training with approximately 10,000 object instances enables models to acquire representations of local geometric structures in 3D shapes. We hypothesize, however, that their representations of 3D global shapes are still limited. To test this hypothesis, we conducted three human experiments systematically manipulating point density and object orientation (Experiment 1), local geometric structure (Experiment 2), and part configuration (Experiment 3). Human performance was stable across conditions in the first two experiments, but declined significantly in the part-scrambled condition of the final experiment. We compared human performance with two types of deep learning architectures: convolution-based models (e.g., DGCNN) and transformer-based models (e.g., Point Transformer). The transformer-based models more closely captured human performance patterns across experimental conditions. Ablation simulations revealed that this advantage is largely driven by progressive downsampling operations that enable hierarchical abstraction of 3D shapes.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1014047","type":"journal-article","created":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T17:48:43Z","timestamp":1773424123000},"page":"e1014047","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Hierarchical abstraction drives human-like 3-D shape processing in deep learning models"],"prefix":"10.1371","volume":"22","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2672-5125","authenticated-orcid":true,"given":"Shuhao","family":"Fu","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Philip J.","family":"Kellman","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hongjing","family":"Lu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"340","published-online":{"date-parts":[[2026,3,13]]},"reference":[{"issue":"4","key":"pcbi.1014047.ref001","doi-asserted-by":"crossref","first-page":"205","DOI":"10.1037\/h0056880","article-title":"The kinetic depth effect","volume":"45","author":"H Wallach","year":"1953","journal-title":"J Exp Psychol"},{"key":"pcbi.1014047.ref002","volume-title":"Vision: A computational investigation into the human representation and processing of visual information","author":"D Marr","year":"1982"},{"issue":"2","key":"pcbi.1014047.ref003","doi-asserted-by":"crossref","first-page":"115","DOI":"10.1037\/0033-295X.94.2.115","article-title":"Recognition-by-components: a theory of human image understanding","volume":"94","author":"I Biederman","year":"1987","journal-title":"Psychol Rev"},{"issue":"3","key":"pcbi.1014047.ref004","doi-asserted-by":"crossref","first-page":"480","DOI":"10.1037\/0033-295X.99.3.480","article-title":"Dynamic binding in a neural network for shape recognition","volume":"99","author":"JE Hummel","year":"1992","journal-title":"Psychol Rev"},{"issue":"4","key":"pcbi.1014047.ref005","doi-asserted-by":"crossref","first-page":"549","DOI":"10.1016\/0042-6989(94)00150-K","article-title":"Object classification for human and ideal observers","volume":"35","author":"Z Liu","year":"1995","journal-title":"Vision Res"},{"issue":"4","key":"pcbi.1014047.ref006","doi-asserted-by":"crossref","first-page":"353","DOI":"10.3758\/BF03202789","article-title":"Perception of three-dimensional form by human infants","volume":"36","author":"PJ Kellman","year":"1984","journal-title":"Percept Psychophys"},{"key":"pcbi.1014047.ref007","first-page":"533","article-title":"Changes in visual object recognition precede the shape bias in early noun learning","volume":"3","author":"M Yee","year":"2012","journal-title":"Front Psychol"},{"issue":"1","key":"pcbi.1014047.ref008","doi-asserted-by":"crossref","first-page":"59","DOI":"10.1016\/0042-6989(91)90074-F","article-title":"Human perception of structure from motion","volume":"31","author":"S Treue","year":"1991","journal-title":"Vision Res"},{"issue":"1","key":"pcbi.1014047.ref009","doi-asserted-by":"crossref","first-page":"139","DOI":"10.1016\/0042-6989(94)E0069-W","article-title":"Structure-from-motion: perceptual evidence for surface interpolation","volume":"35","author":"S Treue","year":"1995","journal-title":"Vision Res"},{"issue":"5","key":"pcbi.1014047.ref010","doi-asserted-by":"crossref","first-page":"508","DOI":"10.1093\/cercor\/13.5.508","article-title":"Processing shape, motion and three-dimensional shape-from-motion in the human cortex","volume":"13","author":"SO Murray","year":"2003","journal-title":"Cereb Cortex"},{"issue":"6","key":"pcbi.1014047.ref011","doi-asserted-by":"crossref","first-page":"1172","DOI":"10.1037\/a0029333","article-title":"A century of Gestalt psychology in visual perception: I. Perceptual grouping and figure-ground organization","volume":"138","author":"J Wagemans","year":"2012","journal-title":"Psychol Bull"},{"issue":"12","key":"pcbi.1014047.ref012","doi-asserted-by":"crossref","first-page":"4338","DOI":"10.1109\/TPAMI.2020.3005434","article-title":"Deep Learning for 3D Point Clouds: A Survey","volume":"43","author":"Y Guo","year":"2021","journal-title":"IEEE Trans Pattern Anal Mach Intell"},{"issue":"8","key":"pcbi.1014047.ref013","doi-asserted-by":"crossref","first-page":"837","DOI":"10.1038\/77754","article-title":"The dynamics of object-selective activation correlate with recognition performance in humans","volume":"3","author":"K Grill-Spector","year":"2000","journal-title":"Nat Neurosci"},{"issue":"6","key":"pcbi.1014047.ref014","doi-asserted-by":"crossref","first-page":"974","DOI":"10.1038\/s41593-019-0392-5","article-title":"Evidence that recurrent circuits are critical to the ventral stream\u2019s execution of core object recognition behavior","volume":"22","author":"K Kar","year":"2019","journal-title":"Nat Neurosci"},{"issue":"12","key":"pcbi.1014047.ref015","doi-asserted-by":"crossref","first-page":"1119","DOI":"10.1016\/j.tics.2022.09.019","article-title":"Does the brain\u2019s ventral visual pathway compute object shape?","volume":"26","author":"V Ayzenberg","year":"2022","journal-title":"Trends in Cognitive Sciences"},{"issue":"5","key":"pcbi.1014047.ref016","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3326362","article-title":"Dynamic Graph CNN for Learning on Point Clouds","volume":"38","author":"Y Wang","year":"2019","journal-title":"ACM Trans Graph"},{"key":"pcbi.1014047.ref017","doi-asserted-by":"crossref","unstructured":"Charles RQ, Su H, Kaichun M, Guibas LJ. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. 77\u201385. https:\/\/doi.org\/10.1109\/cvpr.2017.16","DOI":"10.1109\/CVPR.2017.16"},{"key":"pcbi.1014047.ref018","doi-asserted-by":"crossref","unstructured":"Zhao H, Jiang L, Jia J, Torr PHS, Koltun V. Point Transformer. In: Proceedings of the IEEE\/CVF International Conference on Computer Vision (ICCV), 2021. 16259\u201368.","DOI":"10.1109\/ICCV48922.2021.01595"},{"key":"pcbi.1014047.ref019","unstructured":"Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems, 2012. 1097\u2013105."},{"key":"pcbi.1014047.ref020","article-title":"Very Deep Convolutional Networks for Large-Scale Image Recognition","author":"K Simonyan","year":"2014","journal-title":"CoRR"},{"key":"pcbi.1014047.ref021","doi-asserted-by":"crossref","unstructured":"He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 770\u20138. https:\/\/doi.org\/10.1109\/cvpr.2016.90","DOI":"10.1109\/CVPR.2016.90"},{"key":"pcbi.1014047.ref022","doi-asserted-by":"crossref","first-page":"46","DOI":"10.1016\/j.visres.2020.04.003","article-title":"Local features and global shape information in object classification by deep convolutional neural networks","volume":"172","author":"N Baker","year":"2020","journal-title":"Vision Res"},{"key":"pcbi.1014047.ref023","unstructured":"Geirhos R, Rubisch P, Michaelis C, Bethge M, Wichmann FA, Brendel W. ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness. In: 2018."},{"key":"pcbi.1014047.ref024","article-title":"Deep problems with neural network models of human vision","volume":"46","author":"JS Bowers","year":"2022","journal-title":"Behav Brain Sci"},{"key":"pcbi.1014047.ref025","doi-asserted-by":"crossref","unstructured":"Zhirong Wu, Song S, Khosla A, Fisher Yu, Linguang Zhang, Xiaoou Tang, et al. 3D ShapeNets: A deep representation for volumetric shapes. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015. 1912\u201320. https:\/\/doi.org\/10.1109\/cvpr.2015.7298801","DOI":"10.1109\/CVPR.2015.7298801"},{"key":"pcbi.1014047.ref026","unstructured":"An Tao. dgcnn.pytorch; 2025. [software] Available from: https:\/\/github.com\/antao97\/dgcnn.pytorch"},{"key":"pcbi.1014047.ref027","unstructured":"Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN. Attention is All You Need. Advances in Neural Information Processing Systems. 2017. p. 5998\u20136008."},{"key":"pcbi.1014047.ref028","unstructured":"Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In: 2021."},{"key":"pcbi.1014047.ref029","unstructured":"Qi CR, Yi L, Su H, Guibas LJ. PointNet: Deep Hierarchical Feature Learning on Point Sets in a Metric Space. In: Advances in Neural Information Processing Systems, 2017. 5099\u2013108."},{"key":"pcbi.1014047.ref030","unstructured":"Yang Y o u. Point\u2010Transformers. 2025."},{"issue":"6","key":"pcbi.1014047.ref031","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/2980179.2980238","article-title":"A scalable active framework for region annotation in 3D shape collections","volume":"35","author":"L Yi","year":"2016","journal-title":"ACM Trans Graph"},{"key":"pcbi.1014047.ref032","doi-asserted-by":"crossref","unstructured":"Elhoseiny M, Ghanem B, Hammoud H, Li Y, Mai J, Peng H, et al. PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies. In: Advances in Neural Information Processing Systems 35, 2022. 23192\u2013204. https:\/\/doi.org\/10.52202\/068431-1685","DOI":"10.52202\/068431-1685"},{"issue":"2","key":"pcbi.1014047.ref033","doi-asserted-by":"crossref","first-page":"187","DOI":"10.1007\/s41095-021-0229-5","article-title":"PCT: Point cloud transformer","volume":"7","author":"M-H Guo","year":"2021","journal-title":"Comp Visual Med"},{"key":"pcbi.1014047.ref034","volume-title":"The cradle of knowledge: Development of perception in infancy","author":"PJ Kellman","year":"2000"},{"issue":"9","key":"pcbi.1014047.ref035","doi-asserted-by":"crossref","first-page":"1659","DOI":"10.1016\/S0893-6080(97)00011-7","article-title":"Networks of spiking neurons: The third generation of neural network models","volume":"10","author":"W Maass","year":"1997","journal-title":"Neural Networks"},{"issue":"1","key":"pcbi.1014047.ref036","doi-asserted-by":"crossref","first-page":"6793","DOI":"10.1038\/s41467-024-51110-5","article-title":"High-performance deep spiking neural networks with 0.3 spikes per neuron","volume":"15","author":"A Stanojevic","year":"2024","journal-title":"Nat Commun"},{"issue":"10","key":"pcbi.1014047.ref037","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pcbi.1008215","article-title":"Recurrent neural networks can explain flexible trading of speed and accuracy in biological vision","volume":"16","author":"CJ Spoerer","year":"2020","journal-title":"PLoS Comput Biol"},{"key":"pcbi.1014047.ref038","doi-asserted-by":"crossref","first-page":"176","DOI":"10.1016\/j.conb.2020.11.009","article-title":"Going in circles is the way forward: the role of recurrence in visual inference","volume":"65","author":"RS van Bergen","year":"2020","journal-title":"Curr Opin Neurobiol"},{"key":"pcbi.1014047.ref039","unstructured":"Simonyan K, Zisserman A. Advances in Neural Information Processing Systems. 2014. p. 568\u201376."},{"issue":"6682","key":"pcbi.1014047.ref040","doi-asserted-by":"crossref","first-page":"504","DOI":"10.1126\/science.adi1374","article-title":"Grounded language acquisition through the eyes and ears of a single child","volume":"383","author":"WK Vong","year":"2024","journal-title":"Science"},{"key":"pcbi.1014047.ref041","doi-asserted-by":"crossref","first-page":"305","DOI":"10.1162\/opmi_a_00189","article-title":"Approximating Human-Level 3D Visual Inferences With Deep Neural Networks","volume":"9","author":"TP O\u2019Connell","year":"2025","journal-title":"Open Mind (Camb)"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1014047","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T00:00:00Z","timestamp":1775433600000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1014047","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T17:44:53Z","timestamp":1775497493000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1014047"}},"subtitle":[],"editor":[{"given":"Jian","family":"Liu","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,3,13]]},"references-count":41,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2026,3,13]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1014047","relation":{},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,13]]}}}