{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T15:10:01Z","timestamp":1788793801623,"version":"build-2803163510"},"reference-count":64,"publisher":"Association for Computing Machinery (ACM)","issue":"9","license":[{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Multimedia Comput. Commun. Appl."],"published-print":{"date-parts":[[2026,9,30]]},"abstract":"<jats:p>Recent advancements in 3D Generative Adversarial Networks (3D-GANs) have made significant strides in generating high-fidelity, multi-view-consistent facial images from single-view 2D image databases. These cutting-edge networks integrate the capabilities of StyleGAN and NeRF, offering differentiated control over pose and attributes through distinct subspaces of the latent space. Despite these advancements, a significant challenge remains: the overlapping of these subspaces, which complicates distinct pose and attribute manipulation. Furthermore, previous studies have largely overlooked the crucial task of segregating identity-centric and identity-free information. To address the above issues, our study introduces a model-agnostic, hierarchical disentanglement framework (in terms of underlying 3D architecture), designed to refine the latent space of 3D-GANs. We achieve this by employing an equivariant network to minimize the pose-attribute overlap and a reversible network to delineate identity from identity-free features. Notably, our research marks the first implementation of pose-invariant feature extractors to enable pose equivariance, alongside the innovative use of reversible networks in latent space disentanglement. The proposed latent space, EDSpace, has demonstrated marked improvements over traditional latent spaces in 3D-GANs. It consistently enhances quality metrics across various datasets, methods, and models, underscoring the framework\u2019s robustness and effectiveness in advancing the capabilities of 3D-GANs.<\/jats:p>","DOI":"10.1145\/3827617","type":"journal-article","created":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T15:13:59Z","timestamp":1783005239000},"page":"1-23","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Enhancing the Latent Space of 3D-GANs: A Model-Agnostic Paradigm for Hierarchical Disentanglement"],"prefix":"10.1145","volume":"22","author":[{"ORCID":"https:\/\/orcid.org\/0009-0008-4677-8834","authenticated-orcid":false,"given":"Xiao","family":"Cui","sequence":"first","affiliation":[{"name":"University of Science and Technology of China, Hefei, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-7432-7837","authenticated-orcid":false,"given":"Xuanqing","family":"Cao","sequence":"additional","affiliation":[{"name":"University of Science and Technology of China, Hefei, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7163-6263","authenticated-orcid":false,"given":"Li","family":"Li","sequence":"additional","affiliation":[{"name":"University of Science and Technology of China, Hefei, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1690-9836","authenticated-orcid":false,"given":"Wengang","family":"Zhou","sequence":"additional","affiliation":[{"name":"University of Science and Technology of China, Hefei, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2188-3028","authenticated-orcid":false,"given":"Houqiang","family":"Li","sequence":"additional","affiliation":[{"name":"University of Science and Technology of China, Hefei, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2026,9,7]]},"reference":[{"key":"e_1_3_2_2_2","first-page":"4432","volume-title":"ICCV","author":"Abdal Rameen","year":"2019","unstructured":"Rameen Abdal, Yipeng Qin, and Peter Wonka. 2019. Image2StyleGAN: How to embed images into the StyleGAN latent space? In ICCV, 4432\u20134441."},{"key":"e_1_3_2_3_2","doi-asserted-by":"publisher","DOI":"10.1145\/3447648"},{"key":"e_1_3_2_4_2","first-page":"59","volume-title":"FG","author":"Baltrusaitis Tadas","year":"2018","unstructured":"Tadas Baltrusaitis, Amir Zadeh, Yao Chong Lim, and Louis-Philippe Morency. 2018. OpenFace 2.0: Facial behavior analysis toolkit. In FG. IEEE, 59\u201366."},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.3390\/app152413163"},{"key":"e_1_3_2_6_2","unstructured":"Pablo Calvo-Barl\u00e9s Sergio G. Rodrigo and Luis Mart\u00edn-Moreno. 2025. Learning finite symmetry groups of dynamical systems via equivariance detection. arXiv:2503.03014. Retrieved from https:\/\/arxiv.org\/abs\/2503.03014"},{"key":"e_1_3_2_7_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01565"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1145\/3716389"},{"key":"e_1_3_2_9_2","volume-title":"ICLR","author":"Cui Xiao","year":"2024","unstructured":"Xiao Cui, Qi Sun, Wengang Zhou, and Houqiang Li. 2024. Exploring GPT-4 vision for text-to-image synthesis evaluation. In ICLR. Retrieved from https:\/\/openreview.net\/forum?id=xmQoodG82a"},{"key":"e_1_3_2_10_2","unstructured":"Xiao Cui Wengang Zhou Yang Hu Weilun Wang and Houqiang Li. 2021. Heredity-aware child face image generation with latent space disentanglement. arXiv:2108.11080. Retrieved from https:\/\/arxiv.org\/abs\/2108.11080"},{"key":"e_1_3_2_11_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00482"},{"key":"e_1_3_2_12_2","volume-title":"ICLR","author":"Dey Neel","year":"2021","unstructured":"Neel Dey, Antong Chen, and Soheil Ghafurian. 2021. Group equivariant generative adversarial networks. In ICLR. Retrieved from https:\/\/openreview.net\/forum?id=rgFNuJHHXv"},{"key":"e_1_3_2_13_2","first-page":"2852","volume-title":"CVPR","author":"Dey Rahul","year":"2024","unstructured":"Rahul Dey, Bernhard Egger, Vishnu Naresh Boddeti, Ye Wang, and Tim K. Marks. 2024. CoLa-SDF: Controllable latent StyleSDF for disentangled 3D face generation. In CVPR, 2852\u20132861."},{"issue":"11","key":"e_1_3_2_14_2","doi-asserted-by":"crossref","first-page":"2182","DOI":"10.3390\/electronics13112182","article-title":"RVDR-YOLOv8: A weed target detection model based on improved YOLOv8","volume":"13","author":"Ding Yuanming","year":"2024","unstructured":"Yuanming Ding, Chen Jiang, Lin Song, Fei Liu, and Yunrui Tao. 2024. RVDR-YOLOv8: A weed target detection model based on improved YOLOv8. Electronics 13, 11 (2024), 2182.","journal-title":"Electronics"},{"key":"e_1_3_2_15_2","unstructured":"Laurent Dinh David Krueger and Yoshua Bengio. 2014. Nice: Non-linear independent components estimation. arXiv:1410.8516. Retrieved from https:\/\/arxiv.org\/abs\/1410.8516"},{"key":"e_1_3_2_16_2","unstructured":"Laurent Dinh Jascha Sohl-Dickstein and Samy Bengio. 2016. Density estimation using real NVP. arXiv:1605.08803. Retrieved from https:\/\/arxiv.org\/abs\/1605.08803"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-78389-0_21"},{"key":"e_1_3_2_18_2","unstructured":"Adam Geitgey. 2020. GitHub-Face Recognition. Retrieved from https:\/\/github.com\/ageitgey\/face_recognition"},{"key":"e_1_3_2_19_2","volume-title":"ICLR","author":"Gu Jiatao","year":"2022","unstructured":"Jiatao Gu, Lingjie Liu, Peng Wang, and Christian Theobalt. 2022. StyleNeRF: A style-based 3D aware generator for high-resolution image synthesis. In ICLR. Retrieved from https:\/\/openreview.net\/forum?id=iUuzzTMUw9K"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMM.2025.3542963"},{"key":"e_1_3_2_21_2","unstructured":"Chunming He Rihan Zhang Fengyang Xiao Chengyu Fang Longxiang Tang Yulun Zhang Linghe Kong Deng-Ping Fan Kai Li and Sina Farsiu. 2025. Run: Reversible unfolding network for concealed object segmentation. arXiv:2501.18783. Retrieved from https:\/\/arxiv.org\/abs\/2501.18783"},{"key":"e_1_3_2_22_2","first-page":"20374","volume-title":"CVPR","author":"Hong Yang","year":"2022","unstructured":"Yang Hong, Bo Peng, Haiyao Xiao, Ligang Liu, and Juyong Zhang. 2022. HeadNeRF: A real-time NeRF-based parametric head model. In CVPR, 20374\u201320384."},{"key":"e_1_3_2_23_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.imavis.2024.105369"},{"key":"e_1_3_2_24_2","first-page":"352","volume-title":"CVPR","author":"Jiang Diqiong","year":"2023","unstructured":"Diqiong Jiang, Dan Song, Ruofeng Tong, and Min Tang. 2023. StyleIPSB: Identity-preserving semantic basis of StyleGAN for high fidelity face swapping. In CVPR, 352\u2013361."},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2024.3523422"},{"key":"e_1_3_2_26_2","first-page":"852","volume-title":"NeurIPS","volume":"34","author":"Karras Tero","year":"2021","unstructured":"Tero Karras, Miika Aittala, Samuli Laine, Erik H\u00e4rk\u00f6nen, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. 2021. Alias-free generative adversarial networks. In NeurIPS, Vol. 34, 852\u2013863."},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00453"},{"key":"e_1_3_2_28_2","first-page":"5313","volume-title":"WACV","author":"Katsumata Kai","year":"2024","unstructured":"Kai Katsumata, Duc Minh Vo, Bei Liu, and Hideki Nakayama. 2024. Revisiting latent space of GAN inversion for robust real image editing. In WACV, 5313\u20135322."},{"key":"e_1_3_2_29_2","volume-title":"NeurIPS","volume":"31","author":"Kingma Durk P.","year":"2018","unstructured":"Durk P. Kingma and Prafulla Dhariwal. 2018. Glow: Generative flow with invertible 1x1 convolutions. In NeurIPS, Vol. 31."},{"key":"e_1_3_2_30_2","unstructured":"Theodoros Kouzelis Ioannis Kakogeorgiou Spyros Gidaris and Nikos Komodakis. 2025. EQ-VAE: Equivariance regularized latent space for improved generative image modeling. arXiv:2502.09509. Retrieved from https:\/\/arxiv.org\/abs\/2502.09509"},{"key":"e_1_3_2_31_2","first-page":"2883","volume-title":"AAAI","volume":"38","author":"Lee Chanho","year":"2024","unstructured":"Chanho Lee, Jinsu Son, Hyounguk Shon, Yunho Jeon, and Junmo Kim. 2024. FRED: Towards a full rotation-equivariance in aerial image object detection. In AAAI, Vol. 38, 2883\u20132891."},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1145\/3607535"},{"key":"e_1_3_2_33_2","first-page":"373","volume-title":"ICPR","volume":"15331","author":"Li Peishan","year":"2025","unstructured":"Peishan Li, Yonghong Zhang, Junfei Wang, Guangyi Ma, and Ziwei Yuan. 2025. Adaptive enhanced reversible flow model for remote sensing image super resolution. In ICPR, Vol. 15331. Springer, 373\u2013388."},{"key":"e_1_3_2_34_2","doi-asserted-by":"publisher","DOI":"10.1145\/3634749"},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMM.2023.3271816"},{"issue":"1","key":"e_1_3_2_36_2","doi-asserted-by":"crossref","first-page":"2496406","DOI":"10.1080\/17538947.2025.2496406","article-title":"Triplane generator-based NeRF-GAN framework for single-view ship reconstruction","volume":"18","author":"Liu Tao","year":"2025","unstructured":"Tao Liu, Shiqi Geng, Yucheng Fu, Zhengling Lei, Yuchi Huo, Xiaocai Zhang, Fang Wang, Bing Han, Mei Sha, and Zhongdai Wu. 2025. Triplane generator-based NeRF-GAN framework for single-view ship reconstruction. International Journal of Digital Earth 18, 1 (2025), 2496406.","journal-title":"International Journal of Digital Earth"},{"key":"e_1_3_2_37_2","doi-asserted-by":"crossref","unstructured":"Yidi Liu Dong Li Yuxin Ma Jie Huang Wenlong Zhang Xueyang Fu and Zheng-Jun Zha. 2025. Decouple to reconstruct: High quality UHD restoration via active feature disentanglement and reversible fusion. arXiv:2503.12764. Retrieved from https:\/\/arxiv.org\/abs\/2503.12764","DOI":"10.1109\/ICCV51701.2025.01081"},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.425"},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1145\/3503250"},{"key":"e_1_3_2_40_2","first-page":"737","volume-title":"CVPR","author":"Oh Wonseok","year":"2024","unstructured":"Wonseok Oh and Youngjoo Jo. 2024. From 2D portraits to 3D realities: Advancing GAN inversion for enhanced image synthesis. In CVPR, 737\u2013746."},{"key":"e_1_3_2_41_2","first-page":"13503","volume-title":"CVPR","author":"Or-El Roy","year":"2022","unstructured":"Roy Or-El, Xuan Luo, Mengyi Shan, Eli Shechtman, Jeong Joon Park, and Ira Kemelmacher-Shlizerman. 2022. StyleSDF: High-resolution 3D-consistent image and geometry generation. In CVPR, 13503\u201313513."},{"key":"e_1_3_2_42_2","doi-asserted-by":"publisher","DOI":"10.5244\/C.29.41"},{"key":"e_1_3_2_43_2","doi-asserted-by":"publisher","DOI":"10.1145\/3539661"},{"key":"e_1_3_2_44_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298682"},{"key":"e_1_3_2_45_2","doi-asserted-by":"publisher","DOI":"10.1109\/TCYB.2022.3166761"},{"key":"e_1_3_2_46_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00926"},{"key":"e_1_3_2_47_2","first-page":"24487","volume-title":"NeurIPS","volume":"35","author":"Skorokhodov Ivan","year":"2022","unstructured":"Ivan Skorokhodov, Sergey Tulyakov, Yiqun Wang, and Peter Wonka. 2022. EpiGRAF: Rethinking training of 3D GANs. In NeurIPS, Vol. 35, 24487\u201324501."},{"issue":"7","key":"e_1_3_2_48_2","first-page":"710","article-title":"Group equivariant learning for few-shot image classification","volume":"55","author":"Su Meijuan","year":"2025","unstructured":"Meijuan Su, LeiLei Yan, and Fanzhang Li. 2025. Group equivariant learning for few-shot image classification. Applied Intelligence 55, 7 (2025), 710.","journal-title":"Applied Intelligence"},{"key":"e_1_3_2_49_2","doi-asserted-by":"publisher","DOI":"10.1145\/3706062"},{"key":"e_1_3_2_50_2","doi-asserted-by":"publisher","DOI":"10.1145\/3550454.3555506"},{"key":"e_1_3_2_51_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.02011"},{"key":"e_1_3_2_52_2","unstructured":"Keqiang Sun Shangzhe Wu Zhaoyang Huang Ning Zhang Quan Wang and HongSheng Li. 2022. Controllable 3D face synthesis with conditional generative occupancy fields. arXiv:2206.08361. Retrieved from https:\/\/arxiv.org\/abs\/2206.08361"},{"key":"e_1_3_2_53_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2025.3540593"},{"key":"e_1_3_2_54_2","first-page":"1","volume-title":"ECCV","volume":"15089","author":"Wang Chien-Yao","year":"2024","unstructured":"Chien-Yao Wang, I-Hau Yeh, and Hong-Yuan Mark Liao. 2024. YOLOv9: Learning what you want to learn using programmable gradient information. In ECCV, Vol. 15089. Springer, 1\u201321."},{"key":"e_1_3_2_55_2","doi-asserted-by":"publisher","DOI":"10.1145\/3725735"},{"key":"e_1_3_2_56_2","first-page":"12863","volume-title":"CVPR","author":"Wu Zongze","year":"2021","unstructured":"Zongze Wu, Dani Lischinski, and Eli Shechtman. 2021. StyleSpace analysis: Disentangled controls for StyleGAN image generation. In CVPR, 12863\u201312872."},{"key":"e_1_3_2_57_2","first-page":"1","volume-title":"ICASSP","author":"Xie Jindong","year":"2025","unstructured":"Jindong Xie, Jiachen Liu, Yupei Lin, Jinbao Wang, Xianxu Hou, and Linlin Shen. 2025. High-fidelity editable portrait synthesis with 3D GAN inversion. In ICASSP. IEEE, 1\u20135."},{"key":"e_1_3_2_58_2","doi-asserted-by":"publisher","DOI":"10.1145\/3512797"},{"key":"e_1_3_2_59_2","first-page":"7225","volume-title":"CVPR","author":"Xu Yiran","year":"2024","unstructured":"Yiran Xu, Zhixin Shu, Cameron Smith, Seoung Wug Oh, and Jia-Bin Huang. 2024. In-N-Out: Faithful 3D GAN inversion with volumetric decomposition for face editing. In CVPR, 7225\u20137235."},{"key":"e_1_3_2_60_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00068"},{"key":"e_1_3_2_61_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMM.2023.3322895"},{"key":"e_1_3_2_62_2","doi-asserted-by":"publisher","DOI":"10.1145\/3695463"},{"key":"e_1_3_2_63_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01790"},{"key":"e_1_3_2_64_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-023-01805-x"},{"key":"e_1_3_2_65_2","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2025.3542089"}],"container-title":["ACM Transactions on Multimedia Computing, Communications, and Applications"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3827617","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T14:31:16Z","timestamp":1788791476000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3827617"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,7]]},"references-count":64,"journal-issue":{"issue":"9","published-print":{"date-parts":[[2026,9,30]]}},"alternative-id":["10.1145\/3827617"],"URL":"https:\/\/doi.org\/10.1145\/3827617","relation":{},"ISSN":["1551-6857","1551-6865"],"issn-type":[{"value":"1551-6857","type":"print"},{"value":"1551-6865","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,7]]},"assertion":[{"value":"2025-11-29","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-06-15","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-09-07","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}