Pith. sign in

REVIEW 3 major objections 6 minor 22 references

Latent Radiance Fields with 3D-aware 2D Representations

T0 review · 3 major / 6 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 This paper claims that radiance fields can be built directly in a VAE's 2D latent space and still render photorealistic novel views.

desk verdict A practical three-stage recipe for photorealistic latent-space radiance fields; the cross-dataset decoder-transfer assumption is the soft spot that needs scrutiny. read the letter →

arxiv 2502.09613 v1 pith:CH64OTY6 submitted 2025-02-13 cs.CV

Chaoyi Zhou , Xi Liu , Feng Luo , Siyu Huang This is my paper
classification cs.CV
keywords latentradiancefieldnovelviewsynthesis3DGaussiansplattingvariationalautoencoder3D-aware2Drepresentationscross-datasetgeneralizationtext-to-3Dgenerationfew-shot
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

The paper tries to establish that 3D scenes can be reconstructed and rendered directly in the compressed 2D latent space of a variational autoencoder, rather than in RGB image space, without losing photorealism. It argues that previous latent-space 3D methods fail for two reasons: the latent codes are not 3D-consistent, and the VAE decoder cannot faithfully decode the latents produced by a neural renderer. The proposed fix is a three-stage pipeline: a correspondence-aware autoencoding stage that aligns corresponding latent pixels across views, a latent radiance field that lifts the corrected latents into 3D Gaussians, and a VAE-radiance-field alignment stage that fine-tunes the decoder on latent-image pairs rendered by such fields. If the claim holds, latent-space novel view synthesis at 1/8 resolution becomes a practical alternative to RGB-space rendering, with lower storage and GPU cost and better cross-dataset generalization.

What carries the argument

The central object is the Latent Radiance Field (LRF): a set of 3D Gaussians whose per-Gaussian appearance is a view-dependent spherical-harmonics coefficient living in the VAE's latent space, rasterized by alpha blending into a 2D latent map and then decoded into an image. The carrying mechanism is two alignment stages around this field. A correspondence-aware loss on latent pixels, weighted by camera pose error, removes high-frequency noise and enforces multi-view consistency during encoder fine-tuning. A decoder fine-tuning loss on latent-image pairs rendered from LRFs corrects the distribution shift between rendered and real latent codes. Together they make the latent space 3D-consistent and the decoder radiance-field-compatible, so no per-scene refinement module is needed.

What would settle it

A benchmark could test this directly: train Stage-I and Stage-III exactly as described, then build LRFs on a held-out dataset with a strongly different appearance distribution, such as night-time driving or aerial imagery, and measure decoded novel-view PSNR and LPIPS against RGB-space 3DGS at the same input resolution. If the VAE-RF decoder does not reduce the distribution shift, or per-dataset decoder retraining is needed, the cross-dataset photorealistic claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that radiance field representations constructed from 2D latent representations can yield photorealistic 3D reconstruction, provided 3D awareness is explicitly injected into the latent space. The authors attribute past failures to the loss of multi-view consistency during VAE encoding and to the distribution shift between real latent codes and latents rendered by a radiance field. Their three-stage solution addresses both: the correspondence-aware autoencoding loss makes the latent space geometrically consistent; the latent radiance field (LRF) represents scenes as 3D Gaussians carrying spherical-harmonics latent codes; and the VAE-RF alignment fine-tunes the decoder on latent-image pairs rendered by LRFs. The experiments report that this pipeline outperforms prior latent 3D reconstruction methods and low-resolution image-space baselines on novel view synthesis, three-view few-shot synthesis, and text-to-3D generation, across object-centric, indoor, and unbounded outdoor datasets.

Load-bearing premise

The load-bearing premise is that the decoder alignment step, trained once on latent-image pairs rendered from DL3DV-10K scenes, transfers to latent radiance fields built on unseen datasets with different appearance and camera statistics.

Editorial extensions

If this is right

  • Latent-space novel view synthesis can reach photorealistic quality at 1/8 input resolution, with the reported efficiency gains: about 3.9 minutes of training, 1 GB GPU usage, and 96 MB storage per scene versus 5.9 minutes, 3 GB, and 200 MB for image-space 3DGS.
  • The fine-tuned VAE encoder can be dropped into existing text-to-3D pipelines without retraining the diffusion model, and the aligned decoder improves the image quality of generated objects.
  • On three-view novel view synthesis, latent radiance fields improve PSNR and SSIM over image-space 3DGS and Mip-Splatting at the same input resolution.
  • The components are complementary: the ablation shows encoder fine-tuning alone and decoder fine-tuning alone each improve over the unaligned VAE, and both together give the best result, with decoder alignment contributing the larger gain.
  • Because the encoder and decoder are fine-tuned once on a large corpus, the resulting LRF training does not require per-scene refinement layers, making it a drop-in replacement for existing radiance field pipelines.

Reading between the lines

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

  • The decoder-alignment stage suggests a general recipe the paper does not state: any neural field rendered in a latent space can be paired with a decoder fine-tuned on that renderer's own latent-image pairs, so the correction is not specific to VAE or 3DGS.
  • The correspondence-aware loss could be applied to other pretrained visual features, such as self-supervised or vision-language features, before distilling them into 3D fields, though the paper demonstrates it only on a VAE.
  • A testable extension is whether LRF quality degrades gracefully as the latent representation is compressed below four channels, or as scenes move far from the DL3DV-10K distribution, such as night-time, aerial, or medical imagery.
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 / 6 minor

Summary. The paper proposes a three-stage framework for performing 3D reconstruction directly in the 2D latent space of a pretrained VAE. Stage I fine-tunes the VAE encoder with a correspondence-consistency loss derived from COLMAP multi-view matches; Stage II optimizes a set of 3D Gaussians on the resulting 2D latent maps (the latent radiance field, LRF) using rasterization with an l1 plus D-SSIM loss; Stage III fine-tunes the VAE decoder on latent-image pairs rendered by LRFs of DL3DV-10K scenes. Experiments report novel-view synthesis on MVImgNet, NeRF-LLFF, DL3DV-10K, and Mip-NeRF360, few-shot NVS on LLFF, text-to-3D generation examples, ablations, and an efficiency comparison. The central claim is that latent-space radiance field reconstruction, after injecting 3D awareness into the VAE, achieves photorealistic results and outperforms prior latent-space and low-resolution image-space methods, including on out-of-distribution datasets.

Significance. If the claims hold, the paper makes a useful contribution by showing that 3D Gaussian Splatting can be trained on VAE latent maps at 1/8 resolution and still decode to competitive images, without per-scene refinement modules. The paper's strengths include systematic comparisons across four datasets, a two-component ablation in Table 3, an efficiency analysis in Table 5, and explicit visualizations of latent-space high-frequency noise reduction. The main caveat is that the cross-dataset generalization claim rests on a decoder-transfer assumption that is not directly tested, and one of the key latent-space baselines is not fully specified. These gaps affect the strength of the paper's headline claims but are addressable with additional experiments and reporting.

major comments (3)
  1. [Sec. 4.3, Appendix A.2, Table 3] The cross-dataset generalization claim in Sec. 5.1 is supported only by end-to-end numbers in Table 1. Stage III fine-tunes the decoder on latent-image pairs rendered from LRFs of DL3DV-10K training scenes (Appendix A.2), and the same decoder is then applied to the out-of-distribution datasets MVImgNet, NeRF-LLFF, and Mip-NeRF360. The ablation in Table 3 is DL3DV-only and shows that decoder fine-tuning is the larger contributor (21.16 vs. 21.73 vs. 22.45 PSNR), so the OOD gains in Table 1 cannot be attributed to a transferable VAE-RF alignment without an OOD ablation. Please add with/without decoder fine-tuning results on at least one OOD dataset, or provide quantitative evidence that the LRF rendering residual distribution is similar across datasets; otherwise the cross-dataset claim should be scoped to the in-distribution setting.
  2. [Sec. 5.1, Table 1] The 'Latent-NeRF*' baseline is central to the claim of outperforming state-of-the-art latent-space reconstruction, but the adaptation is never specified. The original Latent-NeRF (Metzer et al., 2022) is a per-scene refinement method for text-guided 3D generation rather than a general NVS method, so the reader cannot tell whether the asterisked variant uses per-scene refinement, what latent resolution and training protocol are used, or how the evaluation is performed. Please specify the adaptation in detail, or replace it with a fully specified latent NVS baseline.
  3. [Sec. 5.2, Figs. 5, 7, 8] The text-to-3D results are presented only as qualitative image comparisons, yet the abstract and introduction claim that the method 'outperforms' existing methods in 3D generation. Without quantitative metrics (e.g., CLIP score, user study, or multi-view consistency measure), the text-to-3D claim is not established. Please add quantitative evaluation or explicitly present these results as qualitative demonstrations.
minor comments (6)
  1. [Eq. (6)] In the correspondence loss, the notation uses 'zi' for the corresponding latent pixel in the neighbouring latent Zj; this should be written as 'zj' to avoid confusion.
  2. [Sec. 4.1, Eq. (6) and Appendix A.1] The construction of lambda_ij by normalizing APE values across all image pairs is not clearly motivated; in particular, the text should state whether larger pose errors are intentionally given more weight and how this interacts with the sum over i and j in Eq. (6).
  3. [Appendix A.4, Table 4] Table 4 shows that the fine-tuned VAE reconstruction PSNR drops from 24.59 to 23.25 on DL3DV-10K, which is not 'minimal impact' as stated in the text; this trade-off should be discussed explicitly as a limitation.
  4. [Tables 1, 2, 4, 5] All quantitative tables report point estimates without variance across scenes; adding standard deviations or per-scene scatter would strengthen the claim of consistent improvement, especially for the cross-dataset comparisons.
  5. [Table 5] The efficiency table does not state the GPU model used for the timing and GPU-usage measurements; this should be reported for reproducibility.
  6. [Sec. 1 and Sec. 6] The claim of being the 'first work showing photorealistic latent-space radiance field reconstruction' should be scoped more carefully, since prior latent-space methods such as Latent-NeRF and ED-NeRF also reconstruct images from latent NeRFs, albeit with per-scene refinement; the novel aspect is better stated as the first without per-scene refinement.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the three-stage pipeline is trained and evaluated on disjoint data, and the only self-citation is a non-load-bearing experimental configuration.

full rationale

The derivation chain is empirical rather than definitional. Stage-I fine-tunes the VAE encoder with a correspondence loss computed from COLMAP pairs on DL3DV training scenes. Stage-II fits latent radiance fields to the resulting latent codes. Stage-III fine-tunes the VAE decoder on paired (rendered-latent, ground-truth-image) data generated from LRFs of DL3DV training scenes. Evaluation is then performed on held-out DL3DV test scenes and on out-of-distribution datasets (MVImgNet, NeRF-LLFF, Mip-NeRF360). The VAE-RF alignment decoder is a trained component, not a parameter fitted to the test scenes, so the novel-view outputs are genuine predictions. The cross-dataset transfer of the decoder is an ordinary generalization claim that could have failed and is tested indirectly on OOD data; this is a correctness risk, not circularity. The comparisons are against external baselines (3DGS, Mip-Splatting, Latent-NeRF, Feature-GS) on standard benchmarks, so the central claim is self-contained against external evidence. The only self-citation is to the authors' 3DGS-Enhancer for the few-shot experimental configuration ("We follow the same experimental configurations as in the previous work (Li et al., 2024; Liu et al., 2024)"), which is not load-bearing for the main results. Accordingly, there is no circular step that reduces the predictions to the method's own inputs.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The pipeline depends on standard VAE and 3DGS machinery, on COLMAP correspondence accuracy, on DL3DV representativeness, and on decoder transfer; no new physical entities are introduced. Several hyperparameters (lambda1, lambda2, LRF optimizer settings) are not reported, which weakens reproducibility but not the conceptual contributions.

free parameters (5)
  • lambda1 (correspondence loss weight) = not reported
    Eq. 7 defines LStageI = LVAE + lambda1 * Lcorres + lambda2 * Lreg, but the paper never reports lambda1. It controls how strongly multi-view consistency is imposed on the latent space.
  • lambda2 (regularization weight) = not reported
    Same equation; lambda2 controls how much the fine-tuned latent space stays close to the original VAE. Its value is never given.
  • lambda_train and lambda_novel (decoder alignment weights) = 0.5, 0.5
    Sec. 4.3, Eq. 11 sets both weights to 0.5 to balance training-view and novel-view decoding; these are hand-chosen.
  • KLweight and Dweight for encoder fine-tuning = 1e-6 and 0.5
    Appendix A.3 reports these weights used during Stage-I VAE fine-tuning; they affect image quality and latent geometry and are hand-set.
  • LRF optimizer hyperparameters = not reported
    Sec. 4.2 and Appendix A.3 do not specify learning rates, iterations, or Gaussian initialization for the latent 3DGS optimization; these determine render quality and the decoder alignment data.
assumptions (5)
  • standard math Multi-view epipolar geometry constraint (Eq. 5) applies to the input image sets.
    Used as motivation for the correspondence consistency loss; assumes known and correct camera poses and point correspondences.
  • domain assumption The pretrained LDM VAE and 3DGS formulations are valid when reused in latent space.
    The paper lifts 3DGS to latent codes without re-deriving its properties; any failure of the base models propagates.
  • domain assumption COLMAP-derived correspondences accurately represent true 3D point matches.
    Stage-I pairs are generated by COLMAP with inlier thresholds; incorrect matches would teach the encoder wrong latent associations.
  • domain assumption The DL3DV-10K training set is representative enough for encoder and decoder fine-tuning to transfer to MVImgNet, NeRF-LLFF, and Mip-NeRF360.
    The cross-dataset claim depends entirely on this transfer; the paper does not measure the domain gap quantitatively.
  • domain assumption The fine-tuned VAE latent space remains compatible with Stable Diffusion U-Net without retraining.
    Sec. 5.2 claims diffusion denoising works with fine-tuned latents, but no quantitative fidelity metric for generated assets is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Latent Radiance Fields with 3D-aware 2D Representations." pith.science (2026). https://pith.science/paper/CH64OTY6

@misc{pith2026250209613,
  author       = {Pith},
  title        = {Pith review of: Latent Radiance Fields with 3D-aware 2D Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CH64OTY6}},
  note         = {Machine review of arXiv:2502.09613}
}
read the original abstract

Latent 3D reconstruction has shown great promise in empowering 3D semantic understanding and 3D generation by distilling 2D features into the 3D space. However, existing approaches struggle with the domain gap between 2D feature space and 3D representations, resulting in degraded rendering performance. To address this challenge, we propose a novel framework that integrates 3D awareness into the 2D latent space. The framework consists of three stages: (1) a correspondence-aware autoencoding method that enhances the 3D consistency of 2D latent representations, (2) a latent radiance field (LRF) that lifts these 3D-aware 2D representations into 3D space, and (3) a VAE-Radiance Field (VAE-RF) alignment strategy that improves image decoding from the rendered 2D representations. Extensive experiments demonstrate that our method outperforms the state-of-the-art latent 3D reconstruction approaches in terms of synthesis performance and cross-dataset generalizability across diverse indoor and outdoor scenes. To our knowledge, this is the first work showing the radiance field representations constructed from 2D latent representations can yield photorealistic 3D reconstruction performance.

Figures

Figures reproduced from arXiv: 2502.09613 by the authors.

Figure 1
Figure 1. This work novelly enables the radiance field representations on the latent space of VAE, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An illustration of our pipeline for creating a latent radiance field in conjunction with [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. A visualization of latent spaces of original and our fine-tuned VAEs. Our method ensures [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: A visual comparison of rendering results. Our method can not only render high-quality [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Visual comparison of different text-to-3D generation methods. Our model enables the [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: A qualitative study of the effect of different fine-tuning stages for view synthesis results. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Samples for text-to-3D generation on the image and latent space. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: More samples for text-to-3D generation on the image space. [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: More NVS results on the DL3DV-10K dataset. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: More NVS results on the NeRF-LLFF dataset. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: More NVS results on the Mip-NeRF360 dataset. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: More NVS results on the MVImgNet dataset. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 7 canonical work pages

  1. [1]

    The ability to remain consistency in large view difference is particularly necessary for the outdoor unbounded reconstruction

    Such over- lapping searching strategy ensures our model not only learns from easy and dense correspondence, but also from challenging cases among far-view image pairs, adding great robutness for our model. The ability to remain consistency in large view difference is particularly necessary for the outdoor unbounded reconstruction. Moreover, we set the min...

    work page 2025

  2. [5]

    Semantic abstraction: Open-world 3D scene understanding from 2D vision-language models

    Huy Ha and Shuran Song. Semantic abstraction: Open-world 3D scene understanding from 2D vision-language models. In Proceedings of the 2022 Conference on Robot Learning,

    work page 2022

  3. [6]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll ´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. arXiv:2111.06377,

    arXiv

  4. [7]

    Auto-encoding variational bayes

    Diederik P Kingma. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114,

    arXiv

  5. [9]

    Jiahe Li, Jiawei Zhang, Xiao Bai, Jin Zheng, Xin Ning, Jun Zhou, and Lin Gu

    URL https://arxiv.org/pdf/2205.15585.pdf. Jiahe Li, Jiawei Zhang, Xiao Bai, Jin Zheng, Xin Ning, Jun Zhou, and Lin Gu. Dngaussian: Op- timizing sparse-view 3d gaussian radiance fields with global-local depth normalization. arXiv preprint arXiv:2403.06912,

    arXiv

  6. [10]

    Latent-nerf for shape-guided generation of 3d shapes and textures

    Gal Metzer, Elad Richardson, Or Patashnik, Raja Giryes, and Daniel Cohen-Or. Latent-nerf for shape-guided generation of 3d shapes and textures. arXiv preprint arXiv:2211.07600,

    arXiv

  7. [11]

    arXiv preprint arXiv:2112.03221,

    Oscar Michel, Roi Bar-On, Richard Liu, Sagie Benaim, and Rana Hanocka. arXiv preprint arXiv:2112.03221,

    arXiv

  8. [13]

    ED-NeRF: Efficient Text-Guided Editing of 3D Scene with Latent Space NeRF

    Jangho Park, Gihyun Kwon, and Jong Chul Ye. Ed-nerf: Efficient text-guided editing of 3d scene using latent space nerf. arXiv preprint arXiv:2310.02712,

    work page Pith review arXiv

Show all 22 references
  1. [14]

    Langsplat: 3d lan- guage gaussian splatting

    Minghan Qin, Wanhua Li, Jiawei Zhou, Haoqian Wang, and Hanspeter Pfister. Langsplat: 3d lan- guage gaussian splatting. arXiv preprint arXiv:2312.16084,

    arXiv

  2. [16]

    Ldm3d: Latent diffusion model for 3d

    12 Published as a conference paper at ICLR 2025 Gabriela Ben Melech Stan, Diana Wofk, Scottie Fox, Alex Redden, Will Saxton, Jean Yu, Estelle Aflalo, Shao-Yen Tseng, Fabio Nonato, Matthias Muller, et al. Ldm3d: Latent diffusion model for 3d. arXiv preprint arXiv:2305.10853,

    2025 arXiv

  3. [17]

    Dreamgaussian: Generative gaussian splatting for efficient 3d content creation

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. Dreamgaussian: Generative gaussian splatting for efficient 3d content creation. arXiv preprint arXiv:2309.16653,

    arXiv

  4. [18]

    One-peace: Exploring one general representation model toward unlimited modalities

    Peng Wang, Shijie Wang, Junyang Lin, Shuai Bai, Xiaohuan Zhou, Jingren Zhou, Xinggang Wang, and Chang Zhou. One-peace: Exploring one general representation model toward unlimited modalities. arXiv preprint arXiv:2305.11172,

    arXiv

  5. [22]

    to achieve better 3D generations with complicated text prompts. A.6 E FFICIENCY ANALYSIS Table 5 demonstrates that our method reduces input resolutions, model storage space, and GPU usage for photorealistic NVS, which is particularly useful in cases with limited communication ...

    2025

  6. [2001]

    After obtaining Eij, the APE is calculated as: AP Eij = ∥Eij − I4×4∥F , (12) where I4×4 is the identity matrix F and represents the Frobenius norm

    13 Published as a conference paper at ICLR 2025 A A PPENDIX A.1 D ETAILS OF CALCULATING THE WEIGHT To compute λij, we first calculate the Absolute Pose Error (APE) for each pose pair using the for- mula: Eij = P −1 i Pj, where Pi and Pj are the different camera poses respectiv...

    2025

  7. [2013]

    Decomposing nerf for editing via feature field distillation

    11 Published as a conference paper at ICLR 2025 Sosuke Kobayashi, Eiichi Matsumoto, and Vincent Sitzmann. Decomposing nerf for editing via feature field distillation. In Advances in Neural Information Processing Systems , volume 35,

    2025

  8. [2016]

    Language embedded 3d gaussians for open-vocabulary scene understanding

    Jin-Chuan Shi, Miao Wang, Hao-Bin Duan, and Shao-Hua Guan. Language embedded 3d gaussians for open-vocabulary scene understanding. arXiv preprint arXiv:2311.18482,

    arXiv

  9. [2017]

    Egolifter: Open-world 3d segmentation for egocentric perception

    Qiao Gu, Zhaoyang Lv, Duncan Frost, Simon Green, Julian Straub, and Chris Sweeney. Egolifter: Open-world 3d segmentation for egocentric perception. arXiv preprint arXiv:2403.18118,

    arXiv

  10. [2020]

    From-ground- to-objects: Coarse-to-fine self-supervised monocular depth estimation of dynamic objects with ground contact prior

    Jaeho Moon, Juan Luis Gonzalez Bello, Byeongjun Kwon, and Munchurl Kim. From-ground- to-objects: Coarse-to-fine self-supervised monocular depth estimation of dynamic objects with ground contact prior. arXiv preprint arXiv:2312.10118,

    arXiv

  11. [2021]

    Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields

    Shijie Zhou, Haoran Chang, Sicheng Jiang, Zhiwen Fan, Zehao Zhu, Dejia Xu, Pradyumna Chari, Suya You, Zhangyang Wang, and Achuta Kadambi. Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields. In Proceedings of the IEEE/CVF Conference on Com- put...

    arXiv

  12. [2022]

    Zoedepth: Zero- shot transfer by combining relative and metric depth

    Shariq Farooq Bhat, Reiner Birkl, Diana Wofk, Peter Wonka, and Matthias M¨uller. Zoedepth: Zero- shot transfer by combining relative and metric depth. arXiv preprint arXiv:2302.12288,

    arXiv

  13. [2023]

    Segment any 3d gaussians

    Jiazhong Cen, Jiemin Fang, Chen Yang, Lingxi Xie, Xiaopeng Zhang, Wei Shen, and Qi Tian. Segment any 3d gaussians. arXiv preprint arXiv:2312.00860,

    arXiv

  14. [2024]

    Zhiwen Fan, Peihao Wang, Xinyu Gong, Yifan Jiang, Dejia Xu, and Zhangyang Wang

    URL https://arxiv.org/abs/2309.16585. Zhiwen Fan, Peihao Wang, Xinyu Gong, Yifan Jiang, Dejia Xu, and Zhangyang Wang. Nerf-sos: Any-view self-supervised object segmentation from complex real-world scenes. arXiv e-prints, pp. arXiv–2209,

    arXiv

Pith tools

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