pith. sign in

arxiv: 2607.01290 · v1 · pith:6MOMOPCNnew · submitted 2026-07-01 · 💻 cs.CV

AnchorSplat: Fast and Structure Consistent Detail Synthesis for Gaussian Splatting

Dexu Zhu , Jiangnan Shao , Xiaofeng Wang , Junxian Duan , Jie Cao , Zheng Zhu , Huaibo Huang This is my paper

Pith reviewed 2026-07-03 21:24 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian Splattingdetail synthesissource-free refinementPoint Anchor Mechanismgeometric consistency3DGS-SR benchmarkzero-shot generalizationsingle-pass densification
0
0 comments X

The pith

AnchorSplat refines 3D Gaussian Splatting assets end-to-end in 3D space without any original multi-view images.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

AnchorSplat proposes a deep network that directly processes 3D Gaussian structures to synthesize missing details and reduce texture noise. It avoids the multi-view inconsistencies and high costs of prior 2D image processing approaches by remaining entirely in 3D. The method requires no source images, making it strictly source-free. Its Point Anchor Mechanism maintains geometric consistency through local offset constraints, while a single-pass multiplication replaces iterative densification. Experiments on the new 3DGS-SR benchmark show state-of-the-art quality with speedups up to 100000 times over optimization baselines and strong zero-shot performance on varied data.

Core claim

AnchorSplat is an end-to-end deep network for 3D-native refinement of Gaussian Splatting assets that operates without original multi-view images; the Point Anchor Mechanism enforces geometric consistency via local offset constraints to mitigate ill-posed mapping and gradient issues, while single-pass multiplication replaces iterative densification, delivering state-of-the-art results on the 3DGS-SR dataset at up to 10^5 times the throughput of optimization methods and robust zero-shot generalization.

What carries the argument

The Point Anchor Mechanism, which enforces geometric consistency via local offset constraints on 3D Gaussian points to mitigate ill-posed mapping and gradient confounding.

If this is right

  • Throughput reaches up to 10^5 times faster than traditional optimization-based refinement methods.
  • The single-pass multiplication mechanism eliminates the need for iterative densification steps.
  • Robust zero-shot generalization holds across generative model outputs and real-world scans.
  • The 3DGS-SR benchmark provides the first large-scale evaluation set for source-free 3DGS refinement.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This approach could integrate directly into pipelines that generate Gaussian models from text or single images, enabling immediate refinement without re-capturing data.
  • Single-pass operation may allow on-device or real-time detail enhancement for large scene models where iterative methods are prohibitive.
  • The source-free design opens possibilities for refining proprietary or legacy Gaussian assets where original capture data no longer exists.

Load-bearing premise

Local offset constraints in the Point Anchor Mechanism can enforce geometric consistency across views even when no original multi-view images are available.

What would settle it

Rendering the refined Gaussians from novel viewpoints and observing visible geometric inconsistencies or texture mismatches that exceed those in the input would falsify the consistency claim.

Figures

Figures reproduced from arXiv: 2607.01290 by Dexu Zhu, Huaibo Huang, Jiangnan Shao, Jie Cao, Junxian Duan, Xiaofeng Wang, Zheng Zhu.

Figure 1
Figure 1. Figure 1: Comparison of 3DSR paradigms. (A) depicts the conventional 2D-centric pipeline: the process requires rendering the asset to LR images, applying 2DSR, and then performing 3D reconstruction. This multi-step process is computationally expen￾sive. (B) shows our novel 3D native paradigm: by directly processing the 3D input, we bypass the costly intermediate 2DSR step, thus achieving significantly higher through… view at source ↗
Figure 2
Figure 2. Figure 2: Framework of AnchorSplat. Our model takes a low-quality 3DGS asset as input and first encodes its non-positional attributes into per-point features. (A) The Point Anchor Mechanism imposes a local geometric constraint by defining an anchor box around each input primitive. (B) The 3DGS Decoder then multiplicatively generates K detail-enhanced primitives strictly within this localized anchor space, utilizing … view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison on 3DGS-SR. We select several representative scenes from the 3DGS-SR test set for visual analysis. For image-based methods, the input is the LR image. 2D super-resolution methods yield artifacts and oversmoothing at geometric edges and high-frequency details due to inconsistency; AnchorSplat avoids these phenomena. 5.2 Results and Analysis Main Results on 3DGS-SR Our primary objectiv… view at source ↗
Figure 4
Figure 4. Figure 4: Ablation study visualization. We perform separate ablations on the Point An￾chor Mechanism and the Multiplicative Primitive Factor K. The visual results clearly demonstrate that the Point Anchor Mechanism is crucial for this task, while the gen￾eration of a high Multiplicative Primitive Factor ensures smoother and richer texture details [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Zero-shot generalization on diverse 3D sources. Without any fine-tuning, An￾chorSplat acts as a plug-and-play enhancer for 3D generative models. It successfully sharpens mechanical boundaries for Trellis outputs (A), enriches complex geometric textures for LGM outputs (B), and robustly enhances details in unconstrained, noisy real-world captures (C). orderly, with a near-total loss of high-frequency detail… view at source ↗
read the original abstract

3D Gaussian Splatting (3DGS) has emerged as a powerful representation for high-fidelity rendering. However, existing assets often suffer from quality bottlenecks such as missing details and texture noise. Prior attempts to enhance these assets via 2D image processing introduce multi-view inconsistencies and high computational costs. In this paper, we propose a novel 3D-native refinement paradigm named AnchorSplat. AnchorSplat is an end-to-end deep network operating directly on 3D structures, avoiding the expensive optimization overhead of traditional 3D-2D-3D pipelines. Crucially, AnchorSplat is a strictly source-free solution requiring no original multi-view images. Central to the proposed method is the Point Anchor Mechanism, which enforces geometric consistency via local offset constraints, mitigating ill-posed mapping and gradient confounding. Furthermore, AnchorSplat replaces iterative densification with a single-pass multiplication mechanism. To facilitate research, we construct 3DGS-SR, the first large-scale benchmark for this task. Experiments demonstrate state-of-the-art results on the 3DGS-SR dataset, with throughput up to $10^5$ times faster than optimization methods. Notably, AnchorSplat exhibits robust zero-shot generalization across diverse data distributions, including generative model outputs and real-world scans.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes AnchorSplat, a 3D-native end-to-end deep network for detail synthesis and refinement of 3D Gaussian Splatting (3DGS) assets. It operates strictly source-free (no original multi-view images required at inference), centers on the Point Anchor Mechanism to enforce geometric consistency via local offset constraints, replaces iterative densification with single-pass multiplication, introduces the 3DGS-SR benchmark, and reports SOTA results with up to 10^5 imes faster throughput plus zero-shot generalization.

Significance. If the source-free consistency and speed claims hold under rigorous validation, the work would offer a practical alternative to optimization-heavy or 2D-image-based refinement pipelines for 3DGS, with the new benchmark providing a useful community resource for this task.

major comments (2)
  1. [Abstract / §3] Abstract and §3 (Point Anchor Mechanism): the claim that local offset constraints mitigate ill-posed mapping and gradient confounding without any original multi-view images rests on an unshown invariance property; the skeptic correctly notes that offsets derived solely from potentially noisy input Gaussians could propagate rather than correct inconsistencies, and no equation or ablation demonstrates robustness under zero-shot distribution shift.
  2. [Abstract / Experiments] Abstract and Experiments section: SOTA results and 10^5 imes speedup are asserted without any reported experimental protocol, error bars, dataset statistics, ablation studies, or baseline definitions, so it is impossible to determine whether the data support the central claims.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'throughput up to 10^5 times faster' should name the exact optimization baselines and hardware to allow direct comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and outline the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract / §3] Abstract and §3 (Point Anchor Mechanism): the claim that local offset constraints mitigate ill-posed mapping and gradient confounding without any original multi-view images rests on an unshown invariance property; the skeptic correctly notes that offsets derived solely from potentially noisy input Gaussians could propagate rather than correct inconsistencies, and no equation or ablation demonstrates robustness under zero-shot distribution shift.

    Authors: We agree that the invariance property of the Point Anchor Mechanism was not explicitly derived or ablated in the submitted version. The local offset constraints are intended to limit the mapping to small, structure-preserving adjustments that reduce gradient confounding, but the referee is correct that robustness to noise propagation and zero-shot shifts requires demonstration. We will add the formal equations for the offset constraint and invariance in §3, along with targeted ablations on noisy inputs and distribution shifts, in the revised manuscript. revision: yes

  2. Referee: [Abstract / Experiments] Abstract and Experiments section: SOTA results and 10^5 times speedup are asserted without any reported experimental protocol, error bars, dataset statistics, ablation studies, or baseline definitions, so it is impossible to determine whether the data support the central claims.

    Authors: The Experiments section (§4) and associated tables provide the full protocol, 3DGS-SR dataset statistics, baseline definitions, ablation studies, and error bars supporting the reported speedups and zero-shot results. We will revise the abstract to include explicit references to these elements and ensure all quantitative claims are directly tied to the reported experimental setup. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper presents AnchorSplat as a novel end-to-end deep network operating directly on 3D structures, with the Point Anchor Mechanism and single-pass multiplication introduced as new components. The abstract and description contain no equations or claims that reduce predictions or consistency enforcement to fitted parameters, self-definitions, or self-citation chains. A new benchmark (3DGS-SR) is constructed for evaluation, providing independent external validation. The derivation chain is self-contained with independent technical content and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities beyond the named mechanism; ledger left empty pending full text.

pith-pipeline@v0.9.1-grok · 5777 in / 1076 out tokens · 37140 ms · 2026-07-03T21:24:32.069689+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

47 extracted references · 9 canonical work pages · 1 internal anchor

  1. [1]

    arXiv preprint arXiv:2411.06390 (2024)

    Chen, Y., Mihajlovic, M., Chen, X., Wang, Y., Prokudin, S., Tang, S.: Splat- former: Point transformer for robust 3d gaussian splatting. arXiv preprint arXiv:2411.06390 (2024)

    work page arXiv 2024

  2. [2]

    IEEE transactions on pattern analysis and machine intelligence (2015)

    Dong, C., Loy, C.C., He, K., Tang, X.: Image super-resolution using deep convo- lutional networks. IEEE transactions on pattern analysis and machine intelligence (2015)

    2015

  3. [3]

    IEEE Transactions on Biometrics, Behavior, and Identity Science (2025)

    Duan, J., Liu, S., Hao, Y., Huang, H., He, R.: Dual frequency-guided spatiotempo- ral feature learning for face forgery detection. IEEE Transactions on Biometrics, Behavior, and Identity Science (2025)

    2025

  4. [4]

    Sen- sors (2025)

    Dumic, E., da Silva Cruz, L.A.: Three-dimensional point cloud applications, datasets, and compression methodologies for remote sensing: A meta-survey. Sen- sors (2025)

    2025

  5. [5]

    Advances in neural information processing systems (2024)

    Fan, Z., Zhang, J., Cong, W., Wang, P., Li, R., Wen, K., Zhou, S., Kadambi, A., Wang, Z., Xu, D., et al.: Large spatial model: End-to-end unposed images to semantic 3d. Advances in neural information processing systems (2024)

    2024

  6. [6]

    arXiv preprint arXiv:2404.10318 (2024)

    Feng, X., He, Y., Wang, Y., Yang, Y., Li, W., Chen, Y., Kuang, Z., Fan, J., Jun, Y., et al.: Srgs: Super-resolution 3d gaussian splatting. arXiv preprint arXiv:2404.10318 (2024)

    work page arXiv 2024

  7. [7]

    In: Proceedings of the 32nd ACM International Conference on Multimedia (2024)

    Han, X., Tang, Y., Wang, Z., Li, X.: Mamba3d: Enhancing local features for 3d point cloud analysis via state space model. In: Proceedings of the 32nd ACM International Conference on Multimedia (2024)

    2024

  8. [8]

    IEEE Transactions on Visualization and Computer Graphics (2024)

    Han, Y., Yu, T., Yu, X., Xu, D., Zheng, B., Dai, Z., Yang, C., Wang, Y., Dai, Q.: Super-nerf: View-consistent detail generation for nerf super-resolution. IEEE Transactions on Visualization and Computer Graphics (2024)

    2024

  9. [9]

    In: ACM SIGGRAPH 2024 conference papers (2024)

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2d gaussian splatting for geo- metrically accurate radiance fields. In: ACM SIGGRAPH 2024 conference papers (2024)

    2024

  10. [10]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (2025)

    Huang, Y., Miyazaki, T., Liu, X., Omachi, S.: Infrared image super-resolution: A systematic review and future trends. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (2025)

    2025

  11. [11]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph. (2023) 16 D. Zhu et al

    2023

  12. [12]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition (2016)

    Kim, J., Lee, J.K., Lee, K.M.: Deeply-recursive convolutional network for image super-resolution. In: Proceedings of the IEEE conference on computer vision and pattern recognition (2016)

    2016

  13. [13]

    In: Proceedings of the AAAI Conference on Artificial Intelligence (2025)

    Ko, H.k., Park, D., Park, Y., Lee, B., Han, J., Park, E.: Sequence matters: Harness- ing video models in 3d super-resolution. In: Proceedings of the AAAI Conference on Artificial Intelligence (2025)

    2025

  14. [14]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2024)

    Lee, J.L., Li, C., Lee, G.H.: Disr-nerf: Diffusion-guided view-consistent super- resolution nerf. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2024)

    2024

  15. [15]

    In: Proceedings of the 31st ACM International Conference on Multimedia (2023)

    Li, G., Xing, W., Zhao, L., Lan, Z., Sun, J., Zhang, Z., Zhang, Q., Lin, H., Lin, Z.: Self-reference image super-resolution via pre-trained diffusion large model and window adjustable transformer. In: Proceedings of the 31st ACM International Conference on Multimedia (2023)

    2023

  16. [16]

    Advances in neural information processing systems (2024)

    Liang, D., Zhou, X., Xu, W., Zhu, X., Zou, Z., Ye, X., Tan, X., Bai, X.: Point- mamba: A simple state space model for point cloud analysis. Advances in neural information processing systems (2024)

    2024

  17. [17]

    In: Proceedings of the IEEE/CVF international conference on computer vision (2021)

    Liang, J., Cao, J., Sun, G., Zhang, K., Van Gool, L., Timofte, R.: Swinir: Image restoration using swin transformer. In: Proceedings of the IEEE/CVF international conference on computer vision (2021)

    2021

  18. [18]

    arXiv preprint arXiv:2405.08609 (2024)

    Lin, J.: Dynamic nerf: A review. arXiv preprint arXiv:2405.08609 (2024)

    work page arXiv 2024

  19. [19]

    In: Proceedings of the 33rd ACM International Conference on Multimedia (2025)

    Liu, Y., Pan, J., Li, Y., Dong, Q., Zhu, C., Guo, Y., Wang, F.: Ultravsr: Achiev- ing ultra-realistic video super-resolution with efficient one-step diffusion space. In: Proceedings of the 33rd ACM International Conference on Multimedia (2025)

    2025

  20. [20]

    Current World Models Lack a Persistent State Core

    Lu, J., Zhu, D., Shi, H., Cai, L., Tang, G., Chen, Y., Cao, J., Tang, D., Zhang, Y., Dai, Y., et al.: Current world models lack a persistent state core. arXiv preprint arXiv:2606.20545 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  21. [21]

    Commu- nications of the ACM (2021)

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: Nerf: Representing scenes as neural radiance fields for view synthesis. Commu- nications of the ACM (2021)

    2021

  22. [22]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition (2017)

    Qi, C.R., Su, H., Mo, K., Guibas, L.J.: Pointnet: Deep learning on point sets for 3d classification and segmentation. In: Proceedings of the IEEE conference on computer vision and pattern recognition (2017)

    2017

  23. [23]

    Advances in neural information processing systems (2017)

    Qi, C.R., Yi, L., Su, H., Guibas, L.J.: Pointnet++: Deep hierarchical feature learn- ing on point sets in a metric space. Advances in neural information processing systems (2017)

    2017

  24. [24]

    IEEE transactions on pattern analysis and ma- chine intelligence (2022)

    Saharia, C., Ho, J., Chan, W., Salimans, T., Fleet, D.J., Norouzi, M.: Image super- resolution via iterative refinement. IEEE transactions on pattern analysis and ma- chine intelligence (2022)

    2022

  25. [25]

    In: European Conference on Computer Vision

    Shen, Y., Ceylan, D., Guerrero, P., Xu, Z., Mitra, N.J., Wang, S., Frühstück, A.: Supergaussian: Repurposing video models for 3d super resolution. In: European Conference on Computer Vision. Springer (2024)

    2024

  26. [26]

    Information Fusion (2025)

    Sohail, S.S., Himeur, Y., Kheddar, H., Amira, A., Fadli, F., Atalla, S., Copiaco, A., Mansoor, W.: Advancing 3d point cloud understanding through deep transfer learning: A comprehensive survey. Information Fusion (2025)

    2025

  27. [27]

    In: Proceedings of the Computer Vision and Pattern Recog- nition Conference (2025)

    Wan, Y., Shao, M., Cheng, Y., Zuo, W.: S2gaussian: Sparse-view super-resolution 3d gaussian splatting. In: Proceedings of the Computer Vision and Pattern Recog- nition Conference (2025)

    2025

  28. [28]

    In: Proceedings of the 30th ACM International Conference on Multimedia (2022) AnchorSplat 17

    Wang, C., Wu, X., Guo, Y.C., Zhang, S.H., Tai, Y.W., Hu, S.M.: Nerf-sr: High quality neural radiance fields using supersampling. In: Proceedings of the 30th ACM International Conference on Multimedia (2022) AnchorSplat 17

    2022

  29. [29]

    Advances in Neural Information Processing Systems37, 118883–118906 (2024)

    Wang, H., Cao, J., Liu, J., Zhou, X., Huang, H., He, R.: Hallo3d: Multi-modal hal- lucination detection and mitigation for consistent 3d content generation. Advances in Neural Information Processing Systems37, 118883–118906 (2024)

    2024

  30. [30]

    Advances in Neural Information Processing Systems (2024)

    Wang, X., Li, M., Liu, W., Zhang, H., Hu, S., Zhang, Y., Zhou, Z., Jin, H.: Un- learnable 3d point clouds: Class-wise transformation is all you need. Advances in Neural Information Processing Systems (2024)

    2024

  31. [31]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (2024)

    Wu, G., Yi, T., Fang, J., Xie, L., Zhang, X., Wei, W., Liu, W., Tian, Q., Wang, X.: 4d gaussian splatting for real-time dynamic scene rendering. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (2024)

    2024

  32. [32]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (2024)

    Wu, X., Jiang, L., Wang, P.S., Liu, Z., Liu, X., Qiao, Y., Ouyang, W., He, T., Zhao, H.: Point transformer v3: Simpler faster stronger. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (2024)

    2024

  33. [33]

    Advances in Neural Information Processing Systems (2022)

    Wu, X., Lao, Y., Jiang, L., Liu, X., Zhao, H.: Point transformer v2: Grouped vector attention and partition-based pooling. Advances in Neural Information Processing Systems (2022)

    2022

  34. [34]

    In: European Conference on Computer Vision

    Wu, Z., Wan, Z., Zhang, J., Liao, J., Xu, D.: Rafe: Generative radiance fields restoration. In: European Conference on Computer Vision. Springer (2024)

    2024

  35. [35]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    Xiao, Z., Wang, X.: Event-based video super-resolution via state space models. In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    2025

  36. [36]

    Resplat: Learning recurrent gaussian splats.arXiv preprint arXiv:2510.08575, 2025

    Xu, H., Barath, D., Geiger, A., Pollefeys, M.: Resplat: Learning recurrent gaussian splats. arXiv preprint arXiv:2510.08575 (2025)

    work page arXiv 2025

  37. [37]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    Xu, Y., Park, T., Zhang, R., Zhou, Y., Shechtman, E., Liu, F., Huang, J.B., Liu, D.: Videogigagan: Towards detail-rich video super-resolution. In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    2025

  38. [38]

    Yan, Z., Li, L., Shao, Y., Chen, S., Wu, Z., Hwang, J.N., Zhao, H., Remondino, F.: 3dsceneeditor:Controllable3dsceneeditingwithgaussiansplatting.arXivpreprint arXiv:2412.01583 (2024)

    work page arXiv 2024

  39. [39]

    arXiv preprint arXiv:2406.10111 (2024)

    Yu, X., Zhu, H., He, T., Chen, Z.: Gaussiansr: 3d gaussian super-resolution with 2d diffusion priors. arXiv preprint arXiv:2406.10111 (2024)

    work page arXiv 2024

  40. [40]

    arXiv preprint arXiv:2508.16467 (2025)

    Zeng, H., Bai, Y., Fu, Y.: Arbitrary-scale 3d gaussian super-resolution. arXiv preprint arXiv:2508.16467 (2025)

    work page arXiv 2025

  41. [41]

    arXiv preprint arXiv:2508.03057 (2025)

    Zhang, T., Liang, Z., Wang, B.: A survey of medical point cloud shape learning: Registration, reconstruction and variation. arXiv preprint arXiv:2508.03057 (2025)

    work page arXiv 2025

  42. [42]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (2025)

    Zhang, W., Zhou, J., Geng, H., Zhang, W., Liu, Y.S.: Gap: Gaussianize any point clouds with text guidance. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (2025)

    2025

  43. [43]

    In: Proceed- ings of the IEEE/CVF international conference on computer vision (2021)

    Zhao, H., Jiang, L., Jia, J., Torr, P.H., Koltun, V.: Point transformer. In: Proceed- ings of the IEEE/CVF international conference on computer vision (2021)

    2021

  44. [44]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    Zheng, M., Sun, L., Dong, J., Pan, J.: Efficient video super-resolution for real- time rendering with decoupled g-buffer guidance. In: Proceedings of the Computer Vision and Pattern Recognition Conference (2025)

    2025

  45. [45]

    Advances in Neural Information Processing Systems (2024)

    Zhou, J., Zhang, W., Liu, Y.S.: Diffgs: Functional gaussian splatting diffusion. Advances in Neural Information Processing Systems (2024)

    2024

  46. [46]

    In: 2025 IEEE International Conference on Multimedia and Expo (ICME)

    Zhu, D., Cao, J., Shao, J., Zhang, Z., Duan, J., He, R.: Mtsd: Simple yet effective self-distillation for generalizable deepfake detection. In: 2025 IEEE International Conference on Multimedia and Expo (ICME). pp. 1–6. IEEE (2025)

    2025

  47. [47]

    Pattern recognition (2024)

    Zhu, Q., Fan, L., Weng, N.: Advancements in point cloud data augmentation for deep learning: A survey. Pattern recognition (2024)

    2024