Pith. sign in

REVIEW 3 major objections 5 minor 70 references

GoLF-NRT: Integrating Global Context and Local Geometry for Few-Shot View Synthesis

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A generalizable neural renderer that fuses scene-level context with local epipolar geometry can synthesize novel views from as few as one to three input images, with adaptive sampling based on attention weights.

desk verdict A credible architectural combination for few-shot view synthesis, but the adaptive-sampling derivation is mathematically shaky and the SOTA claim overreaches. read the letter →

arxiv 2505.19813 v1 pith:7EBETJC4 submitted 2025-05-26 cs.CV

You Wang , Li Fang , Hao Zhu , Fei Hu , Long Ye , Zhan Ma This is my paper
classification cs.CV
keywords few-shotviewsynthesisgeneralizableneuralradiancefieldsrenderingtransformerglobalcontextepipolargeometryadaptivesamplingkernelregressionsparseattention
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

GoLF-NRT is a generalizable neural renderer that synthesizes new views of an unseen scene from as few as one to three input images. Its central claim is that local epipolar geometry alone is not enough in the few-shot regime: the renderer also needs a ray-specific global context feature, built by a 3D transformer with sparse attention, that seeds the local feature aggregation. A second claim is that transformer attention weights can be regularized into a volume-rendering probability density by kernel regression, so adaptive depth sampling concentrates on likely surfaces and sharpens boundaries. If these claims hold, practical view synthesis no longer needs dense multi-view capture or per-scene fitting, and one network can serve both sparse and dense inputs.

What carries the argument

The load-bearing mechanism is a global-local fusion loop. Multi-scale features from the input views are encoded by a 3D transformer whose sparse attention mixes intra-view block, intra-view grid, and inter-view interactions, yielding a scene representation $Z_g$; a decoder transformer turns each target ray into a global context feature $F_g$. This feature becomes the initial query of the view transformer instead of a max-pooled token, and the resulting local epipolar feature $F_l$ is concatenated with it into $F_{g-l}$ for the ray transformer. The adaptive sampler converts ray-transformer attention weights $w_i$ into a smooth PDF $p(d) = \sum_i w_i K(d, d_i) / \sum_i K(d, d_i)$ with a Gaussian kernel $K$, so inverse-transform sampling places extra samples near likely surfaces.

What would settle it

On a depth-labeled dataset containing occluded and reflective regions, compare the peak depth of the kernel-regressed PDF with the true surface depth for each ray; if the peaks drift off-surface exactly where attention is distracted and the adaptive-sampling variant renders worse than uniform sampling, the sampling claim is falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes a coarse-to-fine pipeline in which a global context feature, decoded per target ray from a sparse-attention 3D scene representation, is used as the initial query of the view transformer that aggregates features along epipolar lines. The local geometric feature and the global context feature are concatenated and fed to a ray transformer, whose output is decoded to pixel color by an MLP. The paper further claims that replacing element-wise max pooling with this global-context query resolves depth ambiguity and occlusion artifacts that appear at one to three views, and that the adaptive sampling strategy, which converts ray-transformer attention weights into a smooth PDF by Gaussian kernel regression and then uses inverse-transform sampling, improves geometry perception. The experiments reported across three public datasets at 1, 2, 3, and 10 input views are consistent with the claim that this design outperforms prior generalizable renderers, with the largest margins in the few-shot setting.

Load-bearing premise

The method depends on the ray transformer's attention weights being a trustworthy proxy for surface occupancy, so that after kernel regression and inverse-transform sampling the extra samples really concentrate near the true surface rather than at texture or reflection look-alikes.

Editorial extensions

If this is right

  • A single trained network could render a new scene from one, two, or three posed photos, rather than requiring dozens of views and per-scene optimization.
  • The same architecture remains competitive at 10 views, so the global-context mechanism is not purchased at the cost of many-shot performance.
  • The sparse 3D transformer keeps global context cheap: the paper's own ablation reports 1.88 GFLOPs versus 11.67 GFLOPs for full attention.
  • Treating attention weights as a kernel-regressed PDF lets a transformer-based renderer borrow hierarchical volume-sampling ideas, and the reported 0.32 dB gain comes with cleaner depth at object boundaries.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the attention-to-occupancy transfer is what makes adaptive sampling work, the same global-context-seeded query could aid other multi-view correspondence problems, such as stereo matching or keypoint tracking, where ambiguity grows as viewpoint count shrinks.
  • A stress test the paper does not report is transparent, mirror-like, or repetitive-texture scenes, where feature similarity and real surface occupancy should decouple; if adaptive sampling degrades there, the PDF assumption would need an explicit occupancy prior.
  • The global context feature could also condition feed-forward Gaussian-splatting predictors, which currently depend on accurate per-ray depth guesses and would benefit from a coarse scene-level prior.
  • Making the kernel bandwidth adaptive per ray is a cheap, testable extension that could improve the multimodal depth distributions expected for translucent materials.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes GoLF-NRT, a generalizable neural rendering transformer for few-shot view synthesis. The method combines a global context feature obtained from a sparse 3D transformer with local geometric features along epipolar lines, and introduces an adaptive sampling strategy that converts ray-transformer attention weights into a PDF via kernel regression. The authors report experiments on LLFF, Blender, and Shiny datasets with 1, 2, 3, and 10 input views, including ablations on the LLFF 3-view setting.

Significance. If the technical issues are resolved, the paper would make a useful contribution: the global-context-guided aggregation is a plausible extension of GNT, the sparse attention design is efficient, and the authors provide code and extensive comparisons. However, the current formulation of the adaptive sampling PDF is mathematically incorrect as stated, and the 'state-of-the-art across varying numbers of views' claim is not supported by the paper's own tables.

major comments (3)
  1. [Sec. 3.3, Eqs. (5)-(6)] The function p(d) defined in Eq. (5) is not a probability density. The denominator \sum_i K(d,d_i) depends on d, so p(d) is not normalized; it is a weighted average of the attention weights. Inverse transform sampling from p(d) is therefore not well-defined without an additional normalization step, which is not disclosed. Since Table 3 attributes a 0.32 dB improvement to the kernel-regression adaptive sampling, the claimed gain is not grounded by the described method. Please either present a proper kernel density estimate with a normalization constant and explain how the CDF is computed in practice, or show that the released code performs a different operation. Also, the bandwidth h is a free parameter that is never specified or analyzed.
  2. [Sec. 1 and Table 1] The abstract claims 'state-of-the-art performance across varying numbers of input views,' but Table 1 shows that on Blender with 2 and 3 views, GNT reports higher PSNR (23.47 and 25.80) than GoLF-NRT (22.34 and 24.30), and on 1-view LLFF GoLF-NRT ties CaesarNeRF at 18.31 dB. Please qualify the claim (e.g., to specific datasets, metrics, or settings) or explain the discrepancy.
  3. [Sec. 4.3, Table 3] The ablations are reported for a single run with no error bars or significance tests. The improvements from global context (0.44 dB) and kernel regression (0.32 dB) are within the range of typical run-to-run variance in neural rendering, so without multiple seeds or variance estimates the relative importance of the contributions is uncertain.
minor comments (5)
  1. [Tables 1 and 2] Several entries contain duplicated digits, such as '0.9440.9440.944' and '25.8025.8025.80' in Table 1, '0.8800.8800.880' in Table 2, and '24.2024.2024.20' in Table 3; these should be corrected.
  2. [Sec. 3.3] The notation View-Transformer(F^g|{F^i,P^i}) is not fully defined; please specify how the global context feature is used as the initial query in the attention layers.
  3. [Sec. 4.1] The training data description '1023 models from Google Scanned Object [13], RealEstate10K [69], Spaces scenes [14], and 102 real scenes' is ambiguous; please clarify whether 1023 is the total or a per-source count.
  4. [Related Work] The paper cites LVSM in related work but does not compare against it; adding a discussion or comparison would strengthen the claims, especially since LVSM targets sparse views.
  5. [Sec. 3.4, Eq. (7)] The symbol \hat{F} is used inconsistently; earlier the ray representation is \hat{F}^r, so please use consistent notation throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper's claims rest on external benchmark comparisons and architectural design choices, not on self-referential definitions or load-bearing self-citations.

full rationale

GoLF-NRT's central results are empirical comparisons on held-out public datasets (LLFF, Blender, Shiny) against external baselines such as GNT, EVE-NeRF, and CaesarNeRF. The global-context module and the adaptive-sampling strategy are architectural choices: the global context feature is computed from source-view features and used as an initial query, while the kernel-regression PDF in Eq. (5) is derived from attention weights that are network outputs. Neither quantity is defined in terms of the final rendered color or the evaluation metric, and no parameter is fitted to the test sets. No load-bearing self-citation appears in the text: references to GNT, EVE-NeRF, and CaesarNeRF are external baselines, not author-overlapping support, and no uniqueness theorem or prior result is invoked to force the design. The skeptic's concern about Eq. (5) is that the denominator depends on depth and the expression is not a normalized probability density; that is a formal implementation or correctness question, not circularity, because it does not make the predicted rendering equivalent to an input by construction. The claim of state-of-the-art performance is also not fully supported by the paper's own Table 1 in places, since GNT reports higher PSNR than GoLF-NRT on Blender with 2 and 3 views, but this is a factual inconsistency about how the summary claim is aggregated, not a circular derivation. Therefore the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several hand-chosen sampling and bandwidth parameters and on domain assumptions about attention-as-PDF and cross-domain generalization. No formal derivation is provided; the method is empirical and its reported gains are tightly coupled to the experimental configuration.

free parameters (3)
  • Kernel bandwidth h in Eq. (6) = not reported
    Controls the smoothness of the PDF used for adaptive sampling. No value or selection criterion is reported, and the 0.32 dB ablation gain depends on this choice.
  • Coarse and fine sample counts Nc and Nf = 128 coarse, 64 fine
    Hand-chosen sampling budget along each ray. Affects the fidelity of the estimated PDF and the final rendering quality.
  • Training view-pool parameters k and N = k in (1,3), N in (8,12); N=10 at evaluation
    Controls how source-target pairs are constructed from nearby views. Selected empirically following GNT and may influence few-shot results.
assumptions (4)
  • domain assumption Attention weights learned by the ray transformer can be treated as volume-rendering accumulation weights and converted into a PDF for hierarchical sampling.
    Invoked in Sec. 3.3 Adaptive Sampling. The paper itself states that attention weights differ from ordinary PDFs, then assumes kernel regression closes this gap.
  • domain assumption A per-ray global context feature decoded from a coarse scene representation supplies sufficient prior to guide multi-view epipolar aggregation, even from a single view.
    Central to Secs. 3.2 and 3.3, but the paper provides no quantitative isolation of how much information this global feature carries.
  • domain assumption Multi-scale FPN features at 1/8 and 1/4 resolutions capture the global context and local geometry needed for rendering.
    Assumed in Sec. 3.1 without an analysis of whether the chosen scales are optimal or whether simpler single-scale features would work as well.
  • domain assumption Training on Google Scanned Objects, RealEstate10K, Spaces, and handheld captures generalizes to LLFF, Blender, and Shiny evaluation scenes.
    Described in Sec. 4.1. Cross-dataset generalization is assumed and is not guaranteed by the architecture alone.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GoLF-NRT: Integrating Global Context and Local Geometry for Few-Shot View Synthesis." pith.science (2026). https://pith.science/paper/7EBETJC4

@misc{pith2026250519813,
  author       = {Pith},
  title        = {Pith review of: GoLF-NRT: Integrating Global Context and Local Geometry for Few-Shot View Synthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7EBETJC4}},
  note         = {Machine review of arXiv:2505.19813}
}
read the original abstract

Neural Radiance Fields (NeRF) have transformed novel view synthesis by modeling scene-specific volumetric representations directly from images. While generalizable NeRF models can generate novel views across unknown scenes by learning latent ray representations, their performance heavily depends on a large number of multi-view observations. However, with limited input views, these methods experience significant degradation in rendering quality. To address this limitation, we propose GoLF-NRT: a Global and Local feature Fusion-based Neural Rendering Transformer. GoLF-NRT enhances generalizable neural rendering from few input views by leveraging a 3D transformer with efficient sparse attention to capture global scene context. In parallel, it integrates local geometric features extracted along the epipolar line, enabling high-quality scene reconstruction from as few as 1 to 3 input views. Furthermore, we introduce an adaptive sampling strategy based on attention weights and kernel regression, improving the accuracy of transformer-based neural rendering. Extensive experiments on public datasets show that GoLF-NRT achieves state-of-the-art performance across varying numbers of input views, highlighting the effectiveness and superiority of our approach. Code is available at https://github.com/KLMAV-CUC/GoLF-NRT.

Figures

Figures reproduced from arXiv: 2505.19813 by the authors.

Figure 1
Figure 1. The integration of global context and local geome [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of GoLF-NRT: 1) A FPN extracts multi-scale features from input views. 2) Coarse features are encoded to create scene [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Network architecture of Global Context Feature Extrac [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Illustration of how adaptive sampling with kernel regression enhances local geometric perception. (a) Uniform sampling, lacking [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of GoLF-NRT with GNT, EVE-NeRF, and CaesarNeRF using 3 and 2 input views: (a) Trex scene [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison of GoLF-NRT with GNT, EVE-NeRF, and CaesarNeRF using 10 input views. Rows correspond to the [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: Visualization of global and local attentions. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 7
Figure 7. Figure 7: The depth maps from three sampling strategies show that [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

70 extracted references · 49 canonical work pages

  1. [1]

    Where and how: Mitigating con- fusion in neural radiance fields from sparse inputs

    Yanqi Bao, Yuxin Li, Jing Huo, Tianyu Ding, Xinyue Liang, Wenbin Li, and Yang Gao. Where and how: Mitigating con- fusion in neural radiance fields from sparse inputs. InPro- ceedings of the 31st ACM International Conference on Mul- timedia, pages 2180–2188, 2023

    work page 2023

  2. [2]

    Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields

    Jonathan T Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 5855–5864, 2021

    2021

  3. [3]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5470–5479, 2022

    2022

  4. [4]

    Zip-nerf: Anti-aliased grid-based neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Zip-nerf: Anti-aliased grid-based neural radiance fields. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 19697–19705, 2023

    2023

  5. [5]

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction

    David Charatan, Sizhe Lester Li, Andrea Tagliasacchi, and Vincent Sitzmann. pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19457–19467, 2024

    work page 2024

  6. [6]

    Mvsnerf: Fast general- izable radiance field reconstruction from multi-view stereo

    Anpei Chen, Zexiang Xu, Fuqiang Zhao, Xiaoshuai Zhang, Fanbo Xiang, Jingyi Yu, and Hao Su. Mvsnerf: Fast general- izable radiance field reconstruction from multi-view stereo. InProceedings of the IEEE/CVF international conference on computer vision, pages 14124–14133, 2021

    2021

  7. [7]

    Tensorf: Tensorial radiance fields

    Anpei Chen, Zexiang Xu, Andreas Geiger, Jingyi Yu, and Hao Su. Tensorf: Tensorial radiance fields. InEuropean con- ference on computer vision, pages 333–350. Springer, 2022

    2022

  8. [8]

    Explicit correspondence matching for generalizable neural radiance fields.arXiv preprint arXiv:2304.12294, 2023

    Yuedong Chen, Haofei Xu, Qianyi Wu, Chuanxia Zheng, Tat-Jen Cham, and Jianfei Cai. Explicit correspondence matching for generalizable neural radiance fields.arXiv preprint arXiv:2304.12294, 2023

    arXiv 2023

Show all 70 references
  1. [9]

    Mobilenerf: Exploiting the polygon ras- terization pipeline for efficient neural field rendering on mo- bile architectures

    Zhiqin Chen, Thomas Funkhouser, Peter Hedman, and An- drea Tagliasacchi. Mobilenerf: Exploiting the polygon ras- terization pipeline for efficient neural field rendering on mo- bile architectures. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recogn...

    2023

  2. [10]

    Stereo radiance fields (srf): Learning view syn- thesis for sparse views of novel scenes

    Julian Chibane, Aayush Bansal, Verica Lazova, and Gerard Pons-Moll. Stereo radiance fields (srf): Learning view syn- thesis for sparse views of novel scenes. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7911–7920, 2021

    2021

  3. [11]

    Enhancing nerf akin to enhancing llms: Generalizable nerf transformer with mixture-of-view-experts

    Wenyan Cong, Hanxue Liang, Peihao Wang, Zhiwen Fan, Tianlong Chen, Mukund Varma, Yi Wang, and Zhangyang Wang. Enhancing nerf akin to enhancing llms: Generalizable nerf transformer with mixture-of-view-experts. InProceed- ings of the IEEE/CVF International Conference on Com- pu...

    2023

  4. [12]

    Depth-supervised nerf: Fewer views and faster train- ing for free

    Kangle Deng, Andrew Liu, Jun-Yan Zhu, and Deva Ra- manan. Depth-supervised nerf: Fewer views and faster train- ing for free. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12882– 12891, 2022

    2022

  5. [13]

    Google scanned objects: A high- quality dataset of 3d scanned household items

    Laura Downs, Anthony Francis, Nate Koenig, Brandon Kin- man, Ryan Hickman, Krista Reymann, Thomas B McHugh, and Vincent Vanhoucke. Google scanned objects: A high- quality dataset of 3d scanned household items. In2022 In- ternational Conference on Robotics and Automation (ICRA)...

    2022

  6. [14]

    Deepview: View synthesis with learned gra- dient descent

    John Flynn, Michael Broxton, Paul Debevec, Matthew Du- Vall, Graham Fyffe, Ryan Overbeck, Noah Snavely, and Richard Tucker. Deepview: View synthesis with learned gra- dient descent. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2367...

    2019

  7. [15]

    Surfelnerf: Neu- ral surfel radiance fields for online photorealistic reconstruc- tion of indoor scenes

    Yiming Gao, Yan-Pei Cao, and Ying Shan. Surfelnerf: Neu- ral surfel radiance fields for online photorealistic reconstruc- tion of indoor scenes. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 108–118, 2023

    2023

  8. [16]

    Baking neural ra- diance fields for real-time view synthesis

    Peter Hedman, Pratul P Srinivasan, Ben Mildenhall, Jonathan T Barron, and Paul Debevec. Baking neural ra- diance fields for real-time view synthesis. InProceedings of the IEEE/CVF international conference on computer vision, pages 5875–5884, 2021

    2021

  9. [17]

    Tri-miprf: Tri-mip represen- tation for efficient anti-aliasing neural radiance fields

    Wenbo Hu, Yuling Wang, Lin Ma, Bangbang Yang, Lin Gao, Xiao Liu, and Yuewen Ma. Tri-miprf: Tri-mip represen- tation for efficient anti-aliasing neural radiance fields. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 19774–19783, 2023

    2023

  10. [18]

    Local implicit ray function for gener- alizable radiance field representation

    Xin Huang, Qi Zhang, Ying Feng, Xiaoyu Li, Xuan Wang, and Qing Wang. Local implicit ray function for gener- alizable radiance field representation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 97–107, 2023

    2023

  11. [19]

    Putting nerf on a diet: Semantically consistent few-shot view synthesis

    Ajay Jain, Matthew Tancik, and Pieter Abbeel. Putting nerf on a diet: Semantically consistent few-shot view synthesis. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 5885–5894, 2021

    2021

  12. [20]

    Lvsm: A large view synthesis model with minimal 3d inductive bias, 2024

    Haian Jin, Hanwen Jiang, Hao Tan, Kai Zhang, Sai Bi, Tianyuan Zhang, Fujun Luan, Noah Snavely, and Zexiang Xu. Lvsm: A large view synthesis model with minimal 3d inductive bias, 2024

    2024

  13. [21]

    Geonerf: Generalizing nerf with geometry priors

    Mohammad Mahdi Johari, Yann Lepoittevin, and Franc ¸ois Fleuret. Geonerf: Generalizing nerf with geometry priors. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18365–18375, 2022

    2022

  14. [22]

    Ray tracing volume densities.ACM SIGGRAPH computer graphics, 18(3):165– 174, 1984

    James T Kajiya and Brian P V on Herzen. Ray tracing volume densities.ACM SIGGRAPH computer graphics, 18(3):165– 174, 1984

    1984

  15. [23]

    Viewformer: Nerf-free neural rendering from few images using transformers

    Jon ´aˇs Kulh ´anek, Erik Derner, Torsten Sattler, and Robert Babuˇska. Viewformer: Nerf-free neural rendering from few images using transformers. InEuropean Conference on Computer Vision, pages 198–216. Springer, 2022

    2022

  16. [24]

    Gecon- erf: Few-shot neural radiance fields via geometric consis- tency.arXiv preprint arXiv:2301.10941, 2023

    Min-Seop Kwak, Jiuhn Song, and Seungryong Kim. Gecon- erf: Few-shot neural radiance fields via geometric consis- tency.arXiv preprint arXiv:2301.10941, 2023

    2023 arXiv

  17. [25]

    Mine: Towards continuous depth mpi with nerf for novel view synthesis

    Jiaxin Li, Zijian Feng, Qi She, Henghui Ding, Changhu Wang, and Gim Hee Lee. Mine: Towards continuous depth mpi with nerf for novel view synthesis. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 12578–12588, 2021

    2021

  18. [26]

    Feature pyramid networks for object detection.IEEE Computer Society, 2017

    Tsung Yi Lin, Piotr Dollar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection.IEEE Computer Society, 2017

    2017

  19. [27]

    Editing condi- tional radiance fields

    Steven Liu, Xiuming Zhang, Zhoutong Zhang, Richard Zhang, Jun-Yan Zhu, and Bryan Russell. Editing condi- tional radiance fields. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 5773–5783, 2021

    2021

  20. [28]

    Mvsgaussian: Fast generalizable gaussian splatting recon- struction from multi-view stereo

    Tianqi Liu, Guangcong Wang, Shoukang Hu, Liao Shen, Xinyi Ye, Yuhang Zang, Zhiguo Cao, Wei Li, and Ziwei Liu. Mvsgaussian: Fast generalizable gaussian splatting recon- struction from multi-view stereo. InEuropean Conference on Computer Vision, pages 37–53. Springer, 2024

    2024

  21. [29]

    Neural rays for occlusion-aware image-based render- ing

    Yuan Liu, Sida Peng, Lingjie Liu, Qianqian Wang, Peng Wang, Christian Theobalt, Xiaowei Zhou, and Wenping Wang. Neural rays for occlusion-aware image-based render- ing. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 7824–7833, 2022

    2022

  22. [30]

    Local light field fusion: Practical view syn- thesis with prescriptive sampling guidelines.ACM Transac- tions on Graphics (ToG), 38(4):1–14, 2019

    Ben Mildenhall, Pratul P Srinivasan, Rodrigo Ortiz-Cayon, Nima Khademi Kalantari, Ravi Ramamoorthi, Ren Ng, and Abhishek Kar. Local light field fusion: Practical view syn- thesis with prescriptive sampling guidelines.ACM Transac- tions on Graphics (ToG), 38(4):1–14, 2019

    2019

  23. [31]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    2021

  24. [32]

    Entangled view-epipolar information aggregation for generalizable neural radiance fields

    Zhiyuan Min, Yawei Luo, Wei Yang, Yuesong Wang, and Yi Yang. Entangled view-epipolar information aggregation for generalizable neural radiance fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4906–4916, 2024

    2024

  25. [33]

    Reg- nerf: Regularizing neural radiance fields for view synthesis from sparse inputs

    Michael Niemeyer, Jonathan T Barron, Ben Mildenhall, Mehdi SM Sajjadi, Andreas Geiger, and Noha Radwan. Reg- nerf: Regularizing neural radiance fields for view synthesis from sparse inputs. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, ...

    2022

  26. [34]

    Merf: Memory-efficient radiance fields for real- time view synthesis in unbounded scenes.ACM Transactions on Graphics (TOG), 42(4):1–12, 2023

    Christian Reiser, Rick Szeliski, Dor Verbin, Pratul Srini- vasan, Ben Mildenhall, Andreas Geiger, Jon Barron, and Pe- ter Hedman. Merf: Memory-efficient radiance fields for real- time view synthesis in unbounded scenes.ACM Transactions on Graphics (TOG), 42(4):1–12, 2023

    2023

  27. [35]

    Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction

    Jeremy Reizenstein, Roman Shapovalov, Philipp Henzler, Luca Sbordone, Patrick Labatut, and David Novotny. Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction. InProceedings of the IEEE/CVF international conference on computer vis...

    2021

  28. [36]

    Dense depth pri- ors for neural radiance fields from sparse input views

    Barbara Roessle, Jonathan T Barron, Ben Mildenhall, Pratul P Srinivasan, and Matthias Nießner. Dense depth pri- ors for neural radiance fields from sparse input views. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 12892–12901, 2022

    2022

  29. [37]

    Scene representation transformer: Geometry-free novel view syn- thesis through set-latent scene representations

    Mehdi SM Sajjadi, Henning Meyer, Etienne Pot, Urs Bergmann, Klaus Greff, Noha Radwan, Suhani V ora, Mario Luˇci´c, Daniel Duckworth, Alexey Dosovitskiy, et al. Scene representation transformer: Geometry-free novel view syn- thesis through set-latent scene representations. InPr...

    2022

  30. [38]

    Generalizable patch-based neural render- ing

    Mohammed Suhail, Carlos Esteves, Leonid Sigal, and Ameesh Makadia. Generalizable patch-based neural render- ing. InEuropean Conference on Computer Vision, pages 156–174. Springer, 2022

    2022

  31. [39]

    Light field neural rendering

    Mohammed Suhail, Carlos Esteves, Leonid Sigal, and Ameesh Makadia. Light field neural rendering. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8269–8279, 2022

    2022

  32. [40]

    Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction

    Cheng Sun, Min Sun, and Hwann-Tzong Chen. Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5459– 5469, 2022

    2022

  33. [41]

    Fenerf: Face editing in neural radiance fields

    Jingxiang Sun, Xuan Wang, Yong Zhang, Xiaoyu Li, Qi Zhang, Yebin Liu, and Jue Wang. Fenerf: Face editing in neural radiance fields. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 7672–7682, 2022

    2022

  34. [42]

    Recent advances in im- plicit representation-based 3d shape generation.Visual Intel- ligence, 2(1):9, 2024

    Jia-Mu Sun, Tong Wu, and Lin Gao. Recent advances in im- plicit representation-based 3d shape generation.Visual Intel- ligence, 2(1):9, 2024

    2024

  35. [43]

    Splatter image: Ultra-fast single-view 3d recon- struction

    Stanislaw Szymanowicz, Chrisitian Rupprecht, and Andrea Vedaldi. Splatter image: Ultra-fast single-view 3d recon- struction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10208– 10217, 2024

    2024

  36. [44]

    Image denoising by adaptive kernel regression.Circuits Systems & Computers .conference Record.asilomar Conference on, 2005, 2005

    Hiroyuki Takeda, Sina Farsiu, and Peyman Milanfar. Image denoising by adaptive kernel regression.Circuits Systems & Computers .conference Record.asilomar Conference on, 2005, 2005

    2005

  37. [45]

    Ro- bust kernel regression for restoration and reconstruction of images from sparse noisy data.IEEE, 2006

    Hiroyuki Takeda, Sina Farsiu, and Peyman Milanfar. Ro- bust kernel regression for restoration and reconstruction of images from sparse noisy data.IEEE, 2006

    2006

  38. [46]

    Takeda, S

    H. Takeda, S. Farsiu, and P. Milanfar. Kernel regression for image processing and reconstruction.IEEE Transactions on Image Processing, 2007

    2007

  39. [47]

    Grf: Learning a general radi- ance field for 3d representation and rendering

    Alex Trevithick and Bo Yang. Grf: Learning a general radi- ance field for 3d representation and rendering. InProceed- ings of the IEEE/CVF International Conference on Com- puter Vision, pages 15182–15192, 2021

    2021

  40. [48]

    Maxvit: Multi-axis vision transformer.arXiv e- prints, 2022

    Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, and Yinxiao Li. Maxvit: Multi-axis vision transformer.arXiv e- prints, 2022

    2022

  41. [49]

    Is attention all that nerf needs? InThe Eleventh International Confer- ence on Learning Representations, 2022

    Mukund Varma, Peihao Wang, Xuxi Chen, Tianlong Chen, Subhashini Venugopalan, and Zhangyang Wang. Is attention all that nerf needs? InThe Eleventh International Confer- ence on Learning Representations, 2022

    2022

  42. [50]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

    2017

  43. [51]

    Geometry-biased transformers for novel view synthesis.arXiv preprint arXiv:2301.04650, 2023

    Naveen Venkat, Mayank Agarwal, Maneesh Singh, and Shubham Tulsiani. Geometry-biased transformers for novel view synthesis.arXiv preprint arXiv:2301.04650, 2023

    2023 arXiv

  44. [52]

    Clip-nerf: Text-and-image driven manip- ulation of neural radiance fields

    Can Wang, Menglei Chai, Mingming He, Dongdong Chen, and Jing Liao. Clip-nerf: Text-and-image driven manip- ulation of neural radiance fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3835–3844, 2022

    2022

  45. [53]

    Generalizable neural radiance fields for novel view synthesis with transformer.arXiv preprint arXiv:2206.05375, 2022

    Dan Wang, Xinrui Cui, Septimiu Salcudean, and Z Jane Wang. Generalizable neural radiance fields for novel view synthesis with transformer.arXiv preprint arXiv:2206.05375, 2022

    2022 arXiv

  46. [54]

    Sparsenerf: Distilling depth ranking for few-shot novel view synthesis

    Guangcong Wang, Zhaoxi Chen, Chen Change Loy, and Zi- wei Liu. Sparsenerf: Distilling depth ranking for few-shot novel view synthesis. InProceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 9065–9076, 2023

    2023

  47. [55]

    Ibr- net: Learning multi-view image-based rendering

    Qianqian Wang, Zhicheng Wang, Kyle Genova, Pratul P Srinivasan, Howard Zhou, Jonathan T Barron, Ricardo Martin-Brualla, Noah Snavely, and Thomas Funkhouser. Ibr- net: Learning multi-view image-based rendering. InPro- ceedings of the IEEE/CVF conference on computer vision and p...

    2021

  48. [56]

    Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    2004

  49. [57]

    Nex: Real-time view synthesis with neural basis expansion

    Suttisak Wizadwongsa, Pakkapon Phongthawee, Jiraphon Yenphraphai, and Supasorn Suwajanakorn. Nex: Real-time view synthesis with neural basis expansion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 8534–8543, 2021

    2021

  50. [58]

    Neutex: Neural texture mapping for volumetric neural rendering

    Fanbo Xiang, Zexiang Xu, Milos Hasan, Yannick Hold- Geoffroy, Kalyan Sunkavalli, and Hao Su. Neutex: Neural texture mapping for volumetric neural rendering. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7119–7128, 2021

    2021

  51. [59]

    Sinnerf: Training neural radiance fields on complex scenes from a single image

    Dejia Xu, Yifan Jiang, Peihao Wang, Zhiwen Fan, Humphrey Shi, and Zhangyang Wang. Sinnerf: Training neural radiance fields on complex scenes from a single image. InEuropean Conference on Computer Vision, pages 736–753. Springer, 2022

    2022

  52. [60]

    Murf: Multi-baseline radiance fields

    Haofei Xu, Anpei Chen, Yuedong Chen, Christos Sakaridis, Yulun Zhang, Marc Pollefeys, Andreas Geiger, and Fisher Yu. Murf: Multi-baseline radiance fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20041–20050, 2024

    2024

  53. [61]

    Deforming radiance fields with cages

    Tianhan Xu and Tatsuya Harada. Deforming radiance fields with cages. InEuropean Conference on Computer Vision, pages 159–175. Springer, 2022

    2022

  54. [62]

    Contranerf: Gen- eralizable neural radiance fields for synthetic-to-real novel view synthesis via contrastive learning

    Hao Yang, Lanqing Hong, Aoxue Li, Tianyang Hu, Zhen- guo Li, Gim Hee Lee, and Liwei Wang. Contranerf: Gen- eralizable neural radiance fields for synthetic-to-real novel view synthesis via contrastive learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat...

    2023

  55. [63]

    Freenerf: Im- proving few-shot neural rendering with free frequency reg- ularization

    Jiawei Yang, Marco Pavone, and Yue Wang. Freenerf: Im- proving few-shot neural rendering with free frequency reg- ularization. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8254–8263, 2023

    2023

  56. [64]

    Plenoctrees for real-time rendering of neural radiance fields

    Alex Yu, Ruilong Li, Matthew Tancik, Hao Li, Ren Ng, and Angjoo Kanazawa. Plenoctrees for real-time rendering of neural radiance fields. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 5752– 5761, 2021

    2021

  57. [65]

    pixelnerf: Neural radiance fields from one or few images

    Alex Yu, Vickie Ye, Matthew Tancik, and Angjoo Kanazawa. pixelnerf: Neural radiance fields from one or few images. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4578–4587, 2021

    2021

  58. [66]

    Ners: Neural reflectance surfaces for sparse-view 3d reconstruction in the wild.Advances in Neural Informa- tion Processing Systems, 34:29835–29847, 2021

    Jason Zhang, Gengshan Yang, Shubham Tulsiani, and Deva Ramanan. Ners: Neural reflectance surfaces for sparse-view 3d reconstruction in the wild.Advances in Neural Informa- tion Processing Systems, 34:29835–29847, 2021

    2021

  59. [67]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 586–595, 2018

    2018

  60. [68]

    Gps- gaussian: Generalizable pixel-wise 3d gaussian splatting for real-time human novel view synthesis

    Shunyuan Zheng, Boyao Zhou, Ruizhi Shao, Boning Liu, Shengping Zhang, Liqiang Nie, and Yebin Liu. Gps- gaussian: Generalizable pixel-wise 3d gaussian splatting for real-time human novel view synthesis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

    2024

  61. [69]

    Stereo magnification: Learning view synthesis using multiplane images.arXiv preprint arXiv:1805.09817, 2018

    Tinghui Zhou, Richard Tucker, John Flynn, Graham Fyffe, and Noah Snavely. Stereo magnification: Learning view synthesis using multiplane images.arXiv preprint arXiv:1805.09817, 2018

    2018 arXiv

  62. [70]

    Caesarnerf: Calibrated semantic representation for few-shot generalizable neural rendering

    Haidong Zhu, Tianyu Ding, Tianyi Chen, Ilya Zharkov, Ram Nevatia, and Luming Liang. Caesarnerf: Calibrated semantic representation for few-shot generalizable neural rendering. InEuropean Conference on Computer Vision, 2024

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.