Pith. sign in

REVIEW 4 major objections 5 minor 40 references

MutualNeRF: Improve the Performance of NeRF under Limited Samples with Mutual Information Theory

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Mutual information, measured through CLIP semantic distance and camera/color distance, is proposed as a unified metric that guides NeRF sparse view sampling and few-shot view synthesis, with consistent quality improvements over baselines.

desk verdict The experiments are worth a look; the mutual-information theory around them is not. read the letter →

arxiv 2505.11386 v2 pith:6DSO2HFG submitted 2025-05-16 cs.CV

Zifan Wang , Jingwei Li , Yitang Li , Yunze Liu This is my paper · ORCID
classification cs.CV
keywords NeuralRadianceFieldsmutualinformationsparseviewsamplingfew-shotsynthesisactivelearningCLIPsemanticdistancegreedyapproximationregularization
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

This paper aims to show that mutual information can serve as a single, interpretable metric for making NeRF work well when training images are scarce. The authors operationalize mutual information at two levels: a macro level using semantic distance from CLIP embeddings and a micro level using pixel and ray distance driven by camera positions and RGB differences. Under this metric, sparse view sampling becomes a problem of minimizing mutual information between selected views, solved with a greedy look-ahead algorithm they claim is a 2-approximation. Few-shot view synthesis becomes a problem of maximizing mutual information between rendered and training images, implemented as two plug-and-play regularizers. The paper reports that this unified framing improves PSNR, SSIM, and LPIPS over strong baselines on Blender, DTU, and LLFF in both settings.

What carries the argument

The load-bearing object is the mutual information quantity $I(R,R') = H(R) - H(R|R')$, made computable through Assumption 1: $H(R|R')\propto s(R,R')$ and $H(R|R')\propto d(R,R')$, where $s(R,R') = 1 - \cos(f(R), f(R'))$ with $f$ the CLIP image encoder, and $d(R,R')$ is the expected ray-distance, which Lemma 1 rewrites as $T_1 T_2 \|o - o'\|_2^2 + C$ with $T_1,T_2$ fixed integration lengths and $C$ a constant independent of camera positions. Multi-image mutual information is reduced to the maximum pairwise mutual information following the cited multivariate information decomposition. Lemma 3 bounds the rendered-color distance by $3L\|o-o'\| + C$ under an $L$-Lipschitz assumption on the radiance field, which justifies using color differences as the micro regularizer. The greedy look-ahead sampling algorithm, with its claimed 2-approximation guarantee, is the algorithmic consequence of this setup.

What would settle it

One could test Assumption 1 directly: for a fixed scene, estimate $H(R|R')$ from histogram binning of rendered or training images across many viewpoint pairs and compare it with $s(R,R')$ and $d(R,R')$; if the rank correlation between the measured conditional entropy and the sum $s(R,R')+d(R,R')$ is near zero or negative across those pairs, the proportionality assumption collapses and the 2-approximation guarantee loses its basis.

Watch

Extended reading notes

Core claim

The central claim is that the conditional information $H(R|R')$ needed to infer an unknown image $R'$ from a known image $R$ can be expressed by two measurable proxies: $s(R,R')$, the 1-cosine similarity of CLIP features, and $d(R,R')$, a pixel-space distance that Lemma 1 reduces to the squared Euclidean distance between camera positions. On this basis the paper defines mutual information for multiple images as the maximum pairwise mutual information, and uses it as a unified objective: sparse view sampling minimizes redundancy among chosen views with a greedy algorithm and a 2-approximation guarantee, while few-shot training maximizes information flow from training images to rendered images through a macro semantic regularizer and a micro color-distance regularizer. The paper reports consistent gains over ActiveNeRF in the sampling setting and over DietNeRF, RegNeRF, and FreeNeRF in the few-shot setting, with ablations showing both terms contribute and the semantic term contributing more.

Load-bearing premise

The load-bearing premise is Assumption 1 in Section 4, which says the conditional entropy $H(R|R')$ is proportional to the CLIP semantic distance $s(R,R')$ and to the pixel-space distance $d(R,R')$; this proportionality is stated without derivation, and no experiment in the paper checks that either proxy actually tracks conditional entropy.

Editorial extensions

If this is right

  • In sparse view sampling, both orderings of the sequential selection, semantics-then-pose and pose-then-semantics, beat random sampling, furthest-view sampling, and ActiveNeRF on Blender in the paper's two settings, with the pose-then-semantics ordering giving the highest PSNR and semantics-then-pose giving the best LPIPS.
  • In few-shot synthesis, adding the macro and micro regularizers on top of NeRF, DietNeRF, RegNeRF, and FreeNeRF improves PSNR, SSIM, and LPIPS on Blender with 8 views, DTU with 3 views, and LLFF with 3 views, with the semantic term contributing most of the gain.
  • The greedy algorithm reduces per-iteration constraints from $O(N^2)$ to $O(N)$ while keeping a claimed 2-approximation guarantee for the optimal subset of views.
  • Because the regularizers are plug-and-play, the same mutual-information losses can be attached to other radiance field variants without retraining them from scratch.
  • Ablations show that the semantic and pixel terms are complementary: either alone helps, but combining them is better under both task settings.

Reading between the lines

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

  • The paper does not explore this, but if Assumption 1 holds, the same mutual-information acquisition function should transfer to any capture-planning setting where candidate camera poses are known before images exist, such as robot or drone view planning, and the greedy algorithm would offer the same 2-approximation.
  • The ablations place most of the few-shot gain on the macro term, which suggests a direct testable consequence: substituting a stronger or more scene-appropriate semantic encoder than CLIP should improve results further even if the micro term is unchanged.
  • Because Lemma 3 makes the micro term depend on color differences, the framework likely underestimates useful information in scenes with strong appearance changes such as shadows, highlights, or exposure shifts, where close cameras can still produce very different images; testing on such scenes would clarify where the proxy breaks.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. MutualNeRF proposes a general framework, motivated by mutual information, for improving NeRF under limited samples via two mechanisms: sparse view sampling and few-shot view synthesis. The paper models the conditional information H(R|R) between images as proportional to both CLIP semantic distance and a pixel-space distance based on camera position (Assumption 1), then derives a greedy view-selection algorithm with a claimed 2-approximation guarantee and two plug-and-play regularization terms, L_macro (CLIP semantic consistency) and L_micro (color-distance regularization), intended to maximize mutual information between rendered and training views. Experiments on Blender, DTU, and LLFF under active-learning and few-shot settings report consistent improvements in PSNR, SSIM, and LPIPS over several baselines, including ActiveNeRF, DietNeRF, RegNeRF, and FreeNeRF. The abstract and introduction describe the mutual-information metric as 'theoretically robust' and the greedy algorithm as 'near-optimal.' The main contribution of the paper is this unification of view-selection and regularization under an information-theoretic framing, along with the associated experiments.

Significance. If the theoretical framework were sound, the paper would provide a unified and interpretable principle for both sparse view selection and few-shot NeRF regularization, with an approximation guarantee for the greedy selection algorithm. The empirical results are a genuine strength: the tables are internally consistent, the gains over ActiveNeRF in sparse-view sampling and over RegNeRF/FreeNeRF in few-shot synthesis are consistent across datasets and metrics, and the ablations support the usefulness of each proposed term. The paper also explicitly documents comparison limitations with diffusion-based baselines in its limitations section. However, the central formal claims are not currently load-bearing in the way the paper presents them, because the key assumption and the proofs that rest on it have serious gaps. The practical methods may still be valuable as heuristics, but the claimed theoretical support, the 2-approximation guarantee, and the mutual-information interpretation need either rigorous derivation or a substantial repositioning of the paper's claims.

major comments (4)
  1. [Section 4, Assumption 1] Assumption 1 is ill-posed as stated: it asserts both H(R|R) ∝ s(R,R) and H(R|R) ∝ d(R,R), which for a scalar H implies s(R,R) ∝ d(R,R) unless one of the proportionalities is vacuous. The paper never establishes such a relation between CLIP semantic distance and camera-position distance, and the actual algorithms use s and d as separate, sequential criteria rather than as components of a single quantity. Since every theoretical conclusion, including Lemma 2 and the mutual-information interpretation of L_macro and L_micro, depends on this assumption, the theoretical claims are unsupported without a repaired definition and empirical or analytical validation of the proportionality.
  2. [Appendix A.2, proof of Lemma 2] The proof of the claimed 2-approximation transfers a triangle inequality from d to H 'by Assumption 1,' but the proof's triangle inequality step is not valid. The pixel-space distance d is an expectation of squared L2 distances between points on rays, and squared Euclidean distance does not satisfy the triangle inequality. The semantic distance s, defined as 1 minus cosine similarity, is also not a metric and does not satisfy a triangle inequality. Therefore the pigeonhole argument in Cases 1 and 2 cannot establish H(R_i1|R_i2) ≤ δ, and the 2-approximation guarantee is not proven as written.
  3. [Section 6.1, use of Lemma 3] Lemma 3 provides an upper bound ||Ĉ(r)−Ĉ(r)|| ≤ 3L||o−o||+C, but the text immediately concludes that color difference is a lower bound for camera-position difference and therefore a lower bound for pixel-space distance, and that minimizing color difference reduces pixel-space distance. This reverses the logical direction of the bound: an upper bound does not imply that color difference is a lower bound, and minimizing the left-hand side need not decrease the right-hand side. The connection between L_micro and the pixel-space distance d is therefore not established by Lemma 3.
  4. [Section 6.2, Table 4] The claimed distinction from prior work is partly undermined by the fact that L_macro is the DietNeRF consistency loss, as the paper itself notes in the Table 4 caption, and L_micro is a color-variance regularizer. Without a valid derivation from mutual information, the framework reduces to a combination of an existing semantic-consistency loss and a color-distance regularizer. The experiments show that this combination is empirically useful, but the paper's central novelty claim of a unified, theoretically robust mutual-information metric is not supported by the evidence as presented.
minor comments (5)
  1. [Throughout] The notation R and R for known and unknown images is visually indistinguishable in the text and equations; please use distinct symbols such as R_i and R_j or different fonts.
  2. [Introduction] The phrase 'board challenges' appears to be a typo for 'broad challenges.'
  3. [Section 4, Definition 4] The claim that the mutual information of multiple images is the maximum pairwise mutual information, attributed to Williams and Beer, should be stated as a modeling choice or heuristic rather than as a direct consequence of multivariate information decomposition; the paper should clarify that this is not a standard identity.
  4. [Section 7] The limitations section mentions the absence of diffusion-based baselines, which is appropriate, but it does not acknowledge the limitations of Assumption 1 or the proof gaps in Lemma 2 and the usage of Lemma 3; these should be discussed openly in a revision.
  5. [Tables 3 and 4] The tables report average metrics but do not report standard deviations or per-scene breakdowns for the few-shot experiments; since the gains are sometimes small (e.g., +0.50 dB PSNR over FreeNeRF in Table 3), reporting variance would help assess significance.

Circularity Check

3 steps flagged · score 6.0 of 10

Assumption 1 defines H(R|R) as proportional to CLIP and pixel distances, so the MI regularization, sparse-view criterion, and Lemma 2 near-optimality reduce to minimizing those distances; L_macro is explicitly DietNeRF's loss relabeled.

  1. self definitional [Section 4, Assumption 1]
    "We assume the relative information of two images H(R|R) is proportional to the similarity measure and distance measure between two images, that is, H(R|R)∝s(R,R), H(R|R)∝d(R,R)."

    Under Definition 1, maximizing MI is equivalent to minimizing H(R|R) once H(R) is assumed equal for all images. Assumption 1 then defines H(R|R) as proportional to s(R,R) and d(R,R). That makes the few-shot losses L_macro=s and L_micro (=color variance standing in for d) and the sparse-view selection rule (maximize s and d) into the definition of 'mutual information' rather than consequences of it. No entropy is ever estimated, and no experiment tests the proportionality. The simultaneous statement H∝s and H∝d would also require s∝d, which the paper never establishes; the actual sparse-view procedure uses s and d as separate sequential criteria (S→P, P→S). The MI claim is therefore true by stipulation rather than by derivation.

  2. self definitional [Section 5.1, Lemma 2 and Appendix A.2]
    "By Assumption 1 we know that H(R|R)∝ d(R,R). ... Therefore we can get the triangle inequality of H."

    The 2-approximation proof transfers the triangle inequality of d to H solely by invoking Assumption 1. Thus the 'near-optimal' guarantee is a theorem about the chosen pixel-space distance d, not about information-theoretic mutual information. Since s=1−cosine(CLIP) is not a metric (triangle inequality fails) and the implemented selection uses s and d separately (S→P or P→S), the guarantee does not cover the actual algorithm. The result is a tautological consequence of defining H to be proportional to d; it does not validate that greedy selection maximizes mutual information in any independent sense.

1 more flagged steps
  1. renaming known result [Section 6.2, ablations and Table 4]
    "For DietNeRF, the consistency loss actually belongs to theLmacro, so DietNeRF is a degradation of our framework."

    L_macro is defined in Section 6.1 as s(R,R)=1−cos(CLIP(R),CLIP(R)), which is exactly DietNeRF's semantic consistency loss, as the paper's own table indicates by labeling 'NeRF + L_macro' as 'DietNeRF.' L_micro is color variance, a standard regularizer. Calling these terms 'mutual information regularization' renames existing empirical losses rather than deriving a new quantity. The empirical gains in Tables 3 and 4 validate those known heuristics, not a new prediction from information theory; the 'unified MI framework' is a relabeling of DietNeRF's loss plus a color-variance term.

full rationale

The paper's empirical core is honest: Tables 1, 3, and 4 show that adding CLIP-based semantic regularization, color variance, and camera-distance-based view selection improves FreeNeRF/RegNeRF/DietNeRF on standard benchmarks, so the experimental results are independent evidence that these heuristics help. The circular part is the theory wrapper. Mutual information is defined as I(R,R)=H(R)−H(R|R), and the paper assumes H(R) is equal across images, so maximizing MI is equivalent to minimizing H(R|R). Assumption 1 then sets H(R|R) proportional to the CLIP semantic distance s(R,R) and the pixel/camera distance d(R,R); these are exactly the quantities entering L_macro, L_micro, and the greedy sampling rule. Consequently, every 'mutual information' statement is true by construction rather than by measurement, and the same assumption is what lets the Lemma 2 proof import d's triangle inequality into H. The paper also explicitly identifies L_macro with DietNeRF's consistency loss, confirming that the unified MI framework renames a known semantic-consistency regularizer plus a color-variance term. The simultaneous proportionality H∝s and H∝d is itself ill-posed unless s∝d, which is never shown, and the sequential S→P/P→S selection never defines a single scalar H. The paper's limitations section flags only the missing diffusion baselines, not the unverified Assumption 1 that carries the entire theoretical narrative. Because the empirical gains are real and independent of the self-definitional theory, the circularity is partial rather than total: score 6.

Assumptions & free parameters 1 free parameters · 4 assumptions · 1 invented entities

The paper's theory rests almost entirely on Assumption 1, that H(R|R) is proportional to CLIP and camera or color distances, which is asserted rather than derived. The practical method uses no fitted balance weights, but relies on a hand-chosen subset size in the sequential view selection. No new physical entities are introduced; the only invented entity is the operationalized mutual information metric itself, which has no independent evidence.

free parameters (1)
  • Semantic-stage subset size in S to P selection = 20
    In Ours (S to P), the first stage keeps the 20 candidate views with highest semantic distance before ranking by camera distance. This number is chosen by hand and is not swept or justified.
assumptions (4)
  • ad hoc to paper H(R|R) is proportional to s(R,R) and to d(R,R)
    Section 4 Assumption 1 states the proportionality without derivation. The entire MI formulation for both tasks follows from it.
  • domain assumption All images have equal inherent information H(R)
    Section 4 uses equal H(R) to convert MI maximization into conditional entropy maximization. This is not validated for scenes with varying viewpoint content.
  • domain assumption Multivariate mutual information equals the maximum pairwise MI
    Definition 4 relies on Williams and Beer to decompose MI as a max over pairs. This is one particular partial-information-decomposition choice, not a standard identity.
  • ad hoc to paper Conditional entropy H(R_i|R_j) satisfies a triangle inequality
    Lemma 2's proof uses a triangle inequality for H inherited from the distance proxy d. No information-theoretic basis is given for transferring metric properties to conditional entropy.
invented entities (1)
  • Operationalized image-level mutual information I(R,R)
    purpose: Unify sparse view sampling and few-shot regularization under a single objective
    The paper defines MI via Assumption 1 as a proxy combination of CLIP and pixel distances. True MI is never estimated, so the entity has no falsifiable handle beyond the heuristic losses themselves.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MutualNeRF: Improve the Performance of NeRF under Limited Samples with Mutual Information Theory." pith.science (2026). https://pith.science/paper/6DSO2HFG

@misc{pith2026250511386,
  author       = {Pith},
  title        = {Pith review of: MutualNeRF: Improve the Performance of NeRF under Limited Samples with Mutual Information Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6DSO2HFG}},
  note         = {Machine review of arXiv:2505.11386}
}
read the original abstract

This paper introduces MutualNeRF, a framework enhancing Neural Radiance Field (NeRF) performance under limited samples using Mutual Information Theory. While NeRF excels in 3D scene synthesis, challenges arise with limited data and existing methods that aim to introduce prior knowledge lack theoretical support in a unified framework. We introduce a simple but theoretically robust concept, Mutual Information, as a metric to uniformly measure the correlation between images, considering both macro (semantic) and micro (pixel) levels. For sparse view sampling, we strategically select additional viewpoints containing more non-overlapping scene information by minimizing mutual information without knowing ground truth images beforehand. Our framework employs a greedy algorithm, offering a near-optimal solution. For few-shot view synthesis, we maximize the mutual information between inferred images and ground truth, expecting inferred images to gain more relevant information from known images. This is achieved by incorporating efficient, plug-and-play regularization terms. Experiments under limited samples show consistent improvement over state-of-the-art baselines in different settings, affirming the efficacy of our framework.

Figures

Figures reproduced from arXiv: 2505.11386 by the authors.

Figure 1
Figure 1. The overview of MutualNeRF. We introduce a novel and generic NeRF framework, comprehensively in￾tegrating mutual information from macro (semantic space) and micro perspectives (pixel space). This dual-perspective framework adeptly addresses challenges in sparse view sam￾pling and few-shot view synthesis. In this work, we introduce the concept of Mutual Informa￾tion as an interpretable metric to model correlation. Th… view at source ↗
Figure 2
Figure 2. The overview of our framework. First, we leverage mutual information and relative information to quantify the uncertainty in inferring unknown images conditioned on known ones. This involves decomposing the uncertainty into semantic space distance (macro) and pixel space distance (micro). These distances are converted into specific types tailored for quantifying mutual information in different scenarios. In sparse v… view at source ↗
Figure 3
Figure 3. Quantitative comparison in Active Learning settings on Blender. Given limited input views, our strat￾egy can select better candidate views. Our rendered images without excessively blurry boundaries exhibit greater clarity compared to those rendered by ActiveNeRF. the NeRF model with the newly acquired ground-truth im￾ages corresponding to these selected views, then repeat to step 2. Compared to ActiveNeRF, we modify… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on LLFF. Given 3 input views, we show novel views rendered by FreeNeRF and ours Compared with FreeNeRF. FreeNeRF fails to render sharp outlines in some places, but our additional losses can gain a more detailed skeleton structure and better geome…
Figure 5
Figure 5. Figure 5: Example of our results with 3 input views on the DTU dataset. B.2.3 Limitations on baselines FreeNeRF is a strong baseline that achieves state-of-the-art performance compared to methods using priors from diffusion models across many datasets. We get this conclusion fro…
Figure 6
Figure 6. Figure 6: Example of our results with 3 input views on the LLFF dataset [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 24 canonical work pages

  1. [1]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv \'e J \'e gou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. In Proceedings of the IEEE/CVF international conference on computer vision , pages 9650--9660, 2021

    2021

  2. [2]

    Nerdi: Single-view nerf synthesis with language-guided diffusion as general image priors

    Congyue Deng, Chiyu Jiang, Charles R Qi, Xinchen Yan, Yin Zhou, Leonidas Guibas, Dragomir Anguelov, et al. Nerdi: Single-view nerf synthesis with language-guided diffusion as general image priors. arXiv preprint arXiv:2212.03267 , 2022

    arXiv 2022

  3. [3]

    Plenoxels: Radiance fields without neural networks

    Sara Fridovich-Keil, Alex Yu, Matthew Tancik, Qinhong Chen, Benjamin Recht, and Angjoo Kanazawa. Plenoxels: Radiance fields without neural networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5501--5510, 2022

    work page 2022

  4. [4]

    Scalable active learning by approximated error reduction

    Weijie Fu, Meng Wang, Shijie Hao, and Xindong Wu. Scalable active learning by approximated error reduction. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages 1396--1405, 2018

    work page 2018

  5. [5]

    Baking neural radiance fields for real-time view synthesis

    Peter Hedman, Pratul P Srinivasan, Ben Mildenhall, Jonathan T Barron, and Paul Debevec. Baking neural radiance fields for real-time view synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 5875--5884, 2021

    work page 2021

  6. [6]

    Data-efficient image recognition with contrastive predictive coding

    Olivier Henaff. Data-efficient image recognition with contrastive predictive coding. In International conference on machine learning , pages 4182--4192. PMLR, 2020

    work page 2020

  7. [7]

    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. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 5885--5894, 2021

    work page 2021

  8. [8]

    Large scale multi-view stereopsis evaluation

    Rasmus Jensen, Anders Dahl, George Vogiatzis, Engin Tola, and Henrik Aan s. Large scale multi-view stereopsis evaluation. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 406--413, 2014

    work page 2014

Show all 40 references
  1. [9]

    Infonerf: Ray entropy minimization for few-shot neural volume rendering

    Mijeong Kim, Seonguk Seo, and Bohyung Han. Infonerf: Ray entropy minimization for few-shot neural volume rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12912--12921, 2022

    2022

  2. [10]

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In International conference on machine learning , pages 12888--12900. PMLR, 2022

    2022

  3. [11]

    Contrastive multimodal fusion with tupleinfonce

    Yunze Liu, Qingnan Fan, Shanghang Zhang, Hao Dong, Thomas Funkhouser, and Li Yi. Contrastive multimodal fusion with tupleinfonce. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 754--763, 2021

    2021

  4. [12]

    Nerf in the wild: Neural radiance fields for unconstrained photo collections

    Ricardo Martin-Brualla, Noha Radwan, Mehdi SM Sajjadi, Jonathan T Barron, Alexey Dosovitskiy, and Daniel Duckworth. Nerf in the wild: Neural radiance fields for unconstrained photo collections. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio...

    2021

  5. [13]

    Gnerf: Gan-based neural radiance field without posed camera

    Quan Meng, Anpei Chen, Haimin Luo, Minye Wu, Hao Su, Lan Xu, Xuming He, and Jingyi Yu. Gnerf: Gan-based neural radiance field without posed camera. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 6351--6361, 2021

    2021

  6. [14]

    Local light field fusion: Practical view synthesis with prescriptive sampling guidelines

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

    2019

  7. [15]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. In ECCV , 2020

    2020

  8. [16]

    Regnerf: 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. Regnerf: Regularizing neural radiance fields for view synthesis from sparse inputs. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pa...

    2022

  9. [17]

    Representation learning with contrastive predictive coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748 , 2018

    2018 arXiv

  10. [18]

    Activenerf: Learning where to see with uncertainty estimation

    Xuran Pan, Zihang Lai, Shiji Song, and Gao Huang. Activenerf: Learning where to see with uncertainty estimation. In Computer Vision--ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23--27, 2022, Proceedings, Part XXXIII , pages 230--246. Springer, 2022

    2022

  11. [19]

    Nerfies: Deformable neural radiance fields

    Keunhong Park, Utkarsh Sinha, Jonathan T Barron, Sofien Bouaziz, Dan B Goldman, Steven M Seitz, and Ricardo Martin-Brualla. Nerfies: Deformable neural radiance fields. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 5865--5874, 2021

    2021

  12. [20]

    D-nerf: Neural radiance fields for dynamic scenes

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer. D-nerf: Neural radiance fields for dynamic scenes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 10318--10327, 2021

    2021

  13. [21]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning , p...

    2021

  14. [22]

    Common 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. Common objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Visi...

    2021

  15. [23]

    Active learning for convolutional neural networks: A core-set approach

    Ozan Sener and Silvio Savarese. Active learning for convolutional neural networks: A core-set approach. arXiv preprint arXiv:1708.00489 , 2017

    2017 arXiv

  16. [24]

    Active learning literature survey

    Burr Settles. Active learning literature survey. 2009

    2009

  17. [25]

    Neural geometric level of detail: Real-time rendering with implicit 3d shapes

    Towaki Takikawa, Joey Litalien, Kangxue Yin, Karsten Kreis, Charles Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, and Sanja Fidler. Neural geometric level of detail: Real-time rendering with implicit 3d shapes. In Proceedings of the IEEE/CVF Conference on Computer...

    2021

  18. [26]

    Block-nerf: Scalable large scene neural view synthesis

    Matthew Tancik, Vincent Casser, Xinchen Yan, Sabeek Pradhan, Ben Mildenhall, Pratul P Srinivasan, Jonathan T Barron, and Henrik Kretzschmar. Block-nerf: Scalable large scene neural view synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni...

    2022

  19. [27]

    Contrastive multiview coding

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. Contrastive multiview coding. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XI 16 , pages 776--794. Springer, 2020

    2020

  20. [28]

    On mutual information maximization for representation learning

    Michael Tschannen, Josip Djolonga, Paul K Rubenstein, Sylvain Gelly, and Mario Lucic. On mutual information maximization for representation learning. arXiv preprint arXiv:1907.13625 , 2019

    1907 arXiv

  21. [29]

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

    Guangcong Wang, Zhaoxi Chen, Chen Change Loy, and Ziwei Liu. Sparsenerf: Distilling depth ranking for few-shot novel view synthesis. arXiv preprint arXiv:2303.16196 , 2023

    2023 arXiv

  22. [30]

    Nerf--: Neural radiance fields without known camera parameters

    Zirui Wang, Shangzhe Wu, Weidi Xie, Min Chen, and Victor Adrian Prisacariu. Nerf--: Neural radiance fields without known camera parameters. arXiv preprint arXiv:2102.07064 , 2021

    2021 arXiv

  23. [31]

    Nonnegative decomposition of multivariate information

    Paul L Williams and Randall D Beer. Nonnegative decomposition of multivariate information. arXiv preprint arXiv:1004.2515 , 2010

    2010 arXiv

  24. [32]

    Reconfusion: 3d reconstruction with diffusion priors

    Rundi Wu, Ben Mildenhall, Philipp Henzler, Keunhong Park, Ruiqi Gao, Daniel Watson, Pratul P Srinivasan, Dor Verbin, Jonathan T Barron, Ben Poole, et al. Reconfusion: 3d reconstruction with diffusion priors. arXiv preprint arXiv:2312.02981 , 2023

    2023 arXiv

  25. [33]

    Citynerf: Building nerf at city scale

    Yuanbo Xiangli, Linning Xu, Xingang Pan, Nanxuan Zhao, Anyi Rao, Christian Theobalt, Bo Dai, and Dahua Lin. Citynerf: Building nerf at city scale. arXiv preprint arXiv:2112.05504 , 2021

    2021 arXiv

  26. [34]

    Freenerf: Improving few-shot neural rendering with free frequency regularization

    Jiawei Yang, Marco Pavone, and Yue Wang. Freenerf: Improving few-shot neural rendering with free frequency regularization. arXiv preprint arXiv:2303.07418 , 2023

    2023 arXiv

  27. [35]

    A scalable active framework for region annotation in 3d shape collections

    Li Yi, Vladimir G Kim, Duygu Ceylan, I-Chao Shen, Mengyan Yan, Hao Su, Cewu Lu, Qixing Huang, Alla Sheffer, and Leonidas Guibas. A scalable active framework for region annotation in 3d shape collections. ACM Transactions on Graphics (ToG) , 35(6):1--12, 2016

    2016

  28. [36]

    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. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 5752--5761, 2021

    2021

  29. [37]

    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 Vision and Pattern Recognition , pages 4578--4587, 2021

    2021

  30. [38]

    Nerf++: Analyzing and improving neural radiance fields

    Kai Zhang, Gernot Riegler, Noah Snavely, and Vladlen Koltun. Nerf++: Analyzing and improving neural radiance fields. arXiv preprint arXiv:2010.07492 , 2020

    2010 arXiv

  31. [39]

    Sparsefusion: Distilling view-conditioned diffusion for 3d reconstruction

    Zhizhuo Zhou and Shubham Tulsiani. Sparsefusion: Distilling view-conditioned diffusion for 3d reconstruction. arXiv preprint arXiv:2212.00792 , 2022

    2022 arXiv

  32. [40]

    Temporal coherence for active learning in videos

    Javad Zolfaghari Bengar, Abel Gonzalez-Garcia, Gabriel Villalonga, Bogdan Raducanu, Hamed Habibi Aghdam, Mikhail Mozerov, Antonio M Lopez, and Joost Van de Weijer. Temporal coherence for active learning in videos. In Proceedings of the IEEE/CVF International Conference on Comp...

    2019

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.