Pith. sign in

REVIEW 3 major objections 6 minor 3 cited by

HyperGS: Hyperspectral 3D Gaussian Splatting

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that rendering 3D Gaussians inside a per-scene autoencoder's latent space is the most accurate and fastest way to synthesize novel hyperspectral views, outperforming all compared NeRF- and 3DGS-based baselines on real and…

desk verdict First credible 3DGS-for-hyperspectral paper, but the per-scene autoencoder's train/test split is unstated and its held-out error unmeasured, so the headline gains are not yet established. read the letter →

arxiv 2412.12849 v1 pith:HLYSM4IE submitted 2024-12-17 cs.CV

Christopher Thirgood , Oscar Mendez , Erin Chao Ling , Jon Storey , Simon Hadfield This is my paper · ORCID
classification cs.CV
keywords hyperspectralimagingnovelviewsynthesis3DGaussiansplattingautoencoderlatentspaceneuralradiancefieldsview-dependenteffectsbenchmarkstructure-from-motion
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 claims that hyperspectral novel view synthesis is best tackled by rendering 3D Gaussians in a compact learned latent space rather than in the full spectral space. The authors introduce HyperGS, which compresses each scene's high-dimensional spectra with a per-scene autoencoder, performs Gaussian splatting on the resulting latent spectral signatures, and decodes the rendered latent views back into full hyperspectral images. On two real datasets (128 and 141 channels) and a simulated 228-channel room-scale dataset, HyperGS reports the best average PSNR, SSIM, spectral angle, and RMSE among all compared methods, including a 14 dB accuracy improvement over previously published models. The practical significance is a fast, rasterization-based route to material-aware view synthesis that avoids the training instability and slow rendering of NeRF-based spectral methods.

What carries the argument

The load-bearing device is the per-scene convolutional autoencoder. Its encoder maps each pixel's spectrum into a low-dimensional latent code, the 3D Gaussian splatting stage optimizes in that latent space, and its frozen decoder converts every rendered latent image back into a full hyperspectral image, so the autoencoder's reconstruction fidelity bounds the whole system's accuracy. Around that core, the method combines a hash-encoded view-direction MLP that predicts anisotropic spectral and opacity modulations for each Gaussian, a depth-scaled densification score that reduces artifacts from Gaussians near the camera, and a pixel-wise global pruning score based on spectral difference, opacity, and transmittance that retains the top-K Gaussians per pixel. The training loss couples Charbonnier and cosine-similarity terms for spectral quality with SSIM for spatial consistency.

What would settle it

Compute the autoencoder-only reconstruction error on held-out test views by encoding and decoding ground-truth test spectra and measuring PSNR and spectral angle against the originals. If that AE-only error is close to the gap between HyperGS and the best baseline, the reported gains are mostly decoder memorization of the scene rather than 3DGS localization; if it is much smaller, the latent space is not the limiting factor.

Watch

Extended reading notes

Core claim

HyperGS's central claim is that dimensionality, not representation, is what has kept 3D Gaussian splatting out of hyperspectral imaging, and that moving the splatting into a learned latent space removes the obstacle. Each Gaussian stores a latent spectral signature $f_i \in \mathbb{R}^m$, and a hash-encoded MLP $F_v$ predicts view-dependent spectral and opacity modulations $[\tilde{f}_{i,d}, \tilde{\sigma}_{i,d}]$ for each camera direction. The pixel-level latent blend $\hat{C}(p,d) = \sum_{i\in N} T_{i,d}\alpha_i f_i \tilde{f}_{i,d}$ is then decoded by the frozen autoencoder decoder, $C(p,d) = \mathrm{Dec}(\hat{C}(p,d))$, giving the full spectrum. The scene geometry is initialized by running structure-from-motion on a grayscale projection of the hyperspectral images and re-projecting the recovered 3D points into the latent space, with a depth-scaled densification score and pixel-wise global pruning used to keep the point cloud stable. The paper reports that this pipeline achieves the best average quality metrics on the HS-NeRF Bayspec and SOP datasets and on the simulated ScanNet dataset, with the largest relative gains on low-viewpoint, high-channel scenarios.

Load-bearing premise

The load-bearing premise is that the frozen per-scene autoencoder, trained only with per-pixel Huber loss on training-view spectra, preserves the fine spectral detail the 3DGS stage needs; if its latent space smooths away novel-view reflections or narrow spectral features, no downstream rendering can recover them.

Editorial extensions

If this is right

  • Latent-space splatting becomes the natural template for extending 3DGS to any high-dimensional per-pixel modality, since the Gaussian count and optimization cost no longer grow with the channel count.
  • The depth-scaled densification and pixel-wise pruning make the method usable on noisy, low-viewpoint hyperspectral collections, which are typical outside controlled laboratory turntables.
  • Because the renderer is a rasterizer rather than a volume renderer, HyperGS points toward real-time spectral view synthesis once the decoder and MLP are small enough.
  • The paper's benchmark, which adapts NeRF, MipNeRF, TensoRF, and 3DGS baselines to hyperspectral data, provides a common set of evaluation protocols and metrics for future hyperspectral novel view synthesis work.

Reading between the lines

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

  • A direct test of the paper's decomposition would be to report the autoencoder's own reconstruction error on held-out views; if encoding-decoding ground truth spectra already explains most of HyperGS's advantage, the 3DGS stage is contributing less than the headline suggests.
  • The same latent-space trick should transfer to other high-dimensional sensors such as Raman, multispectral lidar, or circular-polarization imaging, where per-pixel signatures are long and view-dependent effects are local.
  • Because the autoencoder is frozen and trained on training-view pixels only, novel-view specular reflections or out-of-distribution spectra are the most likely failure mode; a stress test with glossy or translucent objects would reveal the ceiling.
  • The depth-scaling factor in the densification score is a generic correction for near-camera Gaussians and could be dropped into RGB 3DGS in scenes with large depth variation.
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

3 major / 6 minor

Summary. The paper introduces HyperGS, a method for hyperspectral novel view synthesis (HNVS) that performs 3D Gaussian Splatting in the latent space of a per-scene convolutional autoencoder. The encoder compresses high-dimensional hyperspectral images, the 3DGS stage optimizes latent spectral signatures together with a view-dependent MLP, and the frozen decoder maps latent renders back to full spectra. The authors also introduce an HNVS benchmark with real HS-NeRF datasets and a simulated ScanNet dataset, report consistent first-place rankings across PSNR, SSIM, SAM, and RMSE, and claim a 14 dB accuracy improvement over previously published models.

Significance. If the central claim is supported, HyperGS is a meaningful advance: it is the first successful adaptation of 3DGS to hyperspectral data, and the proposed benchmark with multiple RGB-NVS and HNVS baselines is a useful resource for the community. The per-scene results in Tables 4, 5, and 7 consistently rank HyperGS first, which is a nontrivial empirical signal. The paper also includes extensive ablations of pruning frequency, pruning score functions, and latent dimensionality. However, the evaluation currently leaves a load-bearing ambiguity about the autoencoder's training split and does not isolate the 3DGS stage from the decoder's spectral prior, so the comparative claims are not yet fully established.

major comments (3)
  1. [Sec. 4.1 and Sec. 5] The autoencoder training split is unspecified, which creates a leakage risk for every reported metric. Section 4.1 states only that the AE is 'trained on the pixel level of the scenes dataset,' while the 90% training / 10% test split in Section 5 is introduced for the NVS evaluation and is not explicitly applied to AE training. If test-view pixels participate in AE training, the decoder in Eq. (10) can overfit to held-out spectra and inflate all PSNR, SSIM, SAM, and RMSE numbers in Tables 1, 2, and 6. Please state explicitly that the AE is trained only on the 90% training views, and report AE-only held-out reconstruction error to demonstrate that the decoder does not carry the novel-view spectral fidelity.
  2. [Eq. (10) and Table 3] Even without train/test leakage, the experiments do not isolate the contribution of the 3DGS geometry from the per-scene decoder. The AE is trained with per-pixel Huber loss (Eq. 6) and no view-consistency objective, so the latent space may smooth away fine novel-view spectral structure, such as the specular reflections highlighted in Figure 6, before the 3DGS stage can represent it. The ablation in Table 3 does not include an AE-only baseline or a held-out-view AE reconstruction error. Please add (a) an AE-only held-out evaluation and (b) an ablation that replaces the learned decoder with a fixed linear or raw-spectral decoder of the same latent dimension, so the reader can attribute the gains to the 3DGS stage rather than to decoder interpolation or memorization.
  3. [Abstract and Section 5] The headline '14 dB accuracy improvement' is not reproducible from the reported tables. The per-dataset PSNR gains over HS-NeRF are 7.3 dB (BaySpec, Table 1), 16.1 dB (SOP, Table 2), and 17.8 dB (simulated ScanNet, Table 6); 14 dB appears to be an undefined cross-dataset average. Please define how the aggregate gain is computed, report all per-dataset numbers alongside it, and avoid stating a single global improvement without specifying the comparison method, the dataset mix, and the averaging procedure. This matters because the abstract's central quantitative claim currently cannot be verified from the paper's own tables.
minor comments (6)
  1. [Sec. 1] There is a typo in the first sentence: 'Synethesis' should be 'Synthesis'.
  2. [Eq. (4)] The transmittance equation uses '(1 − σi αj)', which mixes the global opacity σi with the per-Gaussian αj; in standard 3DGS this factor should be (1 − αj). Please check and correct the notation.
  3. [Tables 4 and 5] Several entries in the per-scene tables appear to be formatting errors or copied values: for example, the MipNeRF row for the Rosemary scene lists SAM as '1000*', and the HS-NeRF row for Rosemary appears identical to the Nerfacto row. These errors make per-scene verification difficult and should be cleaned up.
  4. [Table 10] The caption says the latent-space ablation covers 'all three datasets', but the table lists only Bayspec and SOP, and the latent depths do not match the stated 'division of 4 and 6' (141/4 ≈ 35 rather than 36, and 128/6 ≈ 21 rather than 24). Please correct the caption and the reported depths.
  5. [Eq. (17)] The pruning score subtracts an absolute spectral difference from 1 without specifying the normalization of C*. If ground-truth spectra are not in [0,1], the score range and the pruning threshold τp are undefined. Please state the normalization.
  6. [Sec. 8] The simulated ScanNet evaluation does not describe the train/test split or whether the COLMAP poses are estimated from training views only. Please specify the number of training and test views per scene.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the reported NVS numbers come from a held-out view split and no metric reduces by construction to a fitted value; the only self-citation is non-load-bearing and the AE train/test partition is unspecified, which is a leakage risk rather than circularity.

full rationale

Walking the claimed derivation chain, the core claim is that rendering in the latent space of a frozen per-scene autoencoder and decoding via Eq. (10) yields superior HNVS. The autoencoder is trained with the Huber loss of Eq. (6) on 'the pixel level of the scenes dataset' (Sec. 4.1), and the final prediction is C(p,d)=Dec(Ĉ(p,d)) (Eq. 10). This is a standard per-scene representation: the decoder is not optimized on test views and the 90/10 split of Sec. 5 is a genuine held-out view evaluation against external baselines (HS-NeRF, NeRF variants, 3DGS). The paper never states whether AE training was restricted to the same 90% split, and it never reports AE-only held-out reconstruction error, so the decoder ceiling on novel views is unmeasured; that is a correctness/leakage risk, not a demonstrated circular step, because no equation in the paper forces the reported test metrics to equal the AE's training loss. The only self-citation, Ref. [27] (RaspectLoc), supports the general motivation that hyperspectral imaging is useful in robotics and is not load-bearing in the derivation. Densification (Eq. 15), pruning (Eq. 17), and loss (Eq. 11) are all supervised on training views and are not renamed fits of the test quantities. Accordingly, no circular step can be quoted with a specific reduction, and the paper is self-contained against external benchmarks; score reflects only the minor self-citation and the unaddressed AE-split ambiguity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small set of upstream choices. Most consequential is the per-scene autoencoder: it defines the output space through the frozen decoder, its fidelity on held-out views is never measured, and it is trained without any view-consistency objective. Secondary assumptions are the grayscale SfM initialization, the unmodified 3DGS rasterizer, and the synthetic ScanNet spectra used for the room-scale evaluation. Several thresholds and weights (theta_q, beta_field, lambda, beta, tau_p) are left without values, and the configuration was tuned by ablating on the Bayspec dataset, which is also a headline benchmark. No invented physical entities are introduced.

free parameters (6)
  • latent dimensionality (spectral depth divided by 4) = 36 (Bayspec 141 channels), 32 (SOP 128 channels), 57 (ScanNet 228 channels)
    Chosen from ablations in Supplementary Section 9.3; division by 4 was picked for consistent results across datasets.
  • beta_field (depth-scaling denominator in Eq. 16) = not reported
    Tunable parameter in the split score S(gi); no value or sensitivity study is given, so the densification schedule is unrepeatable as published.
  • theta_q (densification split/clone threshold) = not reported
    Threshold in the condition S(gi) > theta_q (Section 4.2.2); the value is inherited from 3DGS practice but not stated for the latent setting.
  • lambda and beta (loss weights in Eq. 11) = not reported
    Weights balancing SSIM against Charbonnier and cosine terms; they directly set the trade-off between spatial and spectral fidelity in every reported metric.
  • top-K pruning rank threshold tau_p = not reported
    Retention rule in Eq. 18; the threshold and the single-pruning-call timing were selected via ablation (Supplementary Section 9.4) and are not quantified.
  • grayscale SfM channel selection = per-scene channel index not reported
    Section 4.2.1 picks the channel with highest foreground intensity variance per scene; the chosen channel is never specified, making initialization unrepeatable.
assumptions (4)
  • domain assumption The per-scene autoencoder latent space faithfully represents the spectral content of held-out novel views, including view-dependent effects such as specular reflections.
    Sections 4.1-4.2 and Eq. 10 decode every latent render with the frozen decoder; AE-only reconstruction error on test views is never reported, so the decoder's ceiling on novel spectra is unquantified.
  • domain assumption COLMAP grayscale SfM from a single selected spectral channel yields camera poses and sparse geometry accurate enough to initialize 3DGS.
    Section 4.2.1 converts HSI to grayscale using one high-variance channel; scenes whose best channel lacks texture would break point matching and the whole pipeline.
  • standard math The differentiable 3DGS rasterizer and optimization from Kerbl et al. [13] behave as published when operating on latent feature vectors instead of RGB colors.
    Equations 1-5 in Section 3 are adopted unchanged; the paper does not re-derive splatting or transmittance gradients.
  • ad hoc to paper Simulated ScanNet scenes, built by pasting downsampled RRUFF Raman spectra onto semantic labels, are a valid proxy for real room-scale hyperspectral scenes.
    Supplementary Section 8 uses piecewise-constant spectral maps with no realistic illumination or sensor noise, which can favor smooth interpolation methods; real room-scale HSI may be harder.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HyperGS: Hyperspectral 3D Gaussian Splatting." pith.science (2026). https://pith.science/paper/HLYSM4IE

@misc{pith2026241212849,
  author       = {Pith},
  title        = {Pith review of: HyperGS: Hyperspectral 3D Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HLYSM4IE}},
  note         = {Machine review of arXiv:2412.12849}
}
read the original abstract

We introduce HyperGS, a novel framework for Hyperspectral Novel View Synthesis (HNVS), based on a new latent 3D Gaussian Splatting (3DGS) technique. Our approach enables simultaneous spatial and spectral renderings by encoding material properties from multi-view 3D hyperspectral datasets. HyperGS reconstructs high-fidelity views from arbitrary perspectives with improved accuracy and speed, outperforming currently existing methods. To address the challenges of high-dimensional data, we perform view synthesis in a learned latent space, incorporating a pixel-wise adaptive density function and a pruning technique for increased training stability and efficiency. Additionally, we introduce the first HNVS benchmark, implementing a number of new baselines based on recent SOTA RGB-NVS techniques, alongside the small number of prior works on HNVS. We demonstrate HyperGS's robustness through extensive evaluation of real and simulated hyperspectral scenes with a 14db accuracy improvement upon previously published models.

Figures

Figures reproduced from arXiv: 2412.12849 by the authors.

Figure 1
Figure 1. This image represents a novel hyperspectral image at the 70th channel of a 141-channel image predicted from the top three [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visual system diagram of our approach. Blue lines indicate the operational flow, while orange lines represent the gradient flow. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Our channel-wise convolutional AE model learns LHSI [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualization of our re-projection protocol for initializ [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Visualization of our pixel-wise pruning. Gaussians with [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Visualisation of the top 4 methods for frame 51 of 359 for the Caladium plant scene from the Bayspec dataset. The top row [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Visualisation of the top 4 methods for frame 31 of 49 for the Tools plant scene from the SOP dataset. The top row shows the 70th [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Three random pixel reconstructions taken from test frame 151 of the bayspec Anacampseros scene. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Three random pixel reconstructions from test frame 31 of the SOP origami scene. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UnMix-NeRF: Spectral Unmixing Meets Neural Radiance Fields

    eess.IV 2025-06 conditional novelty 6.0 of 10

    A NeRF-based framework jointly performs hyperspectral novel view synthesis and unsupervised material segmentation by learning per-point spectral abundances over a global endmember dictionary.

  2. Towards Integrating Multi-Spectral Imaging with Gaussian Splatting

    cs.CV 2025-08 conditional novelty 5.0 of 10

    Jointly optimizing RGB and four additional spectral bands in one 3D Gaussian Splatting model, after an RGB-only warm-up and with spectrum-aware densification, outperforms per-band models and slightly improves RGB via ...

  3. A Survey on 3D Reconstruction Techniques in Plant Phenotyping: From Classical Methods to Neural Radiance Fields (NeRF), 3D Gaussian Splatting (3DGS), and Beyond

    eess.IV 2025-04 conditional novelty 1.0 of 10

    A review of 3D reconstruction techniques for plant phenotyping, comparing classical methods, NeRF, and 3D Gaussian Splatting on methodology, applications, and future directions.

Reference graph

Works this paper leans on

34 extracted references · 27 canonical work pages · cited by 3 Pith papers

  1. [1]

    Barron, B

    J.T. Barron, B. Mildenhall, D. Verbin, P.P. Srinivasan, and P. Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 4, 6

    work page 2022

  2. [2]

    Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P

    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, 2021. 6

    work page 2021

  3. [3]

    S. R. Bul `o, L. Porzi, and P. Kontschieder. Revising densifi- cation in gaussian splatting. ArXiv, 2024. 2

    work page 2024

  4. [4]

    Tensorf: Tensorial radiance fields

    Anpei Chen, Zexiang Xu, Andreas Geiger, Jingyi Yu, and Hao Su. Tensorf: Tensorial radiance fields. In European Conference on Computer Vision (ECCV), 2022. 2, 6

    work page 2022

  5. [5]

    G. Chen, S. K. Narayanan, T. G. Ottou, B. Missaoui, H. Muriki, C. Pradalier, and Y . Chen. Hyperspectral neural ra- diance fields. ArXiv, 2024. 2, 6

    work page 2024

  6. [6]

    Gaussianpro: 3d gaussian splatting with progressive propagation

    Kai Cheng, Xiaoxiao Long, Kaizhi Yang, Yao Yao, Wei Yin, Yuexin Ma, Wenping Wang, and Xuejin Chen. Gaussianpro: 3d gaussian splatting with progressive propagation. arXiv preprint arXiv:2402.14650, 2024. 2

    arXiv 2024

  7. [7]

    Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner

    Angela Dai, Angel X. Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proc. Computer Vision and Pattern Recognition (CVPR), IEEE, 2017. 6, 1

    work page 2017

  8. [8]

    Dellaert

    F. Dellaert. NeRF explosion 2020. https://dellaert. github.io/NeRF/, 2020. 2

    work page 2020

Show all 34 references
  1. [9]

    Z. Fan, K. Wang, K. Wen, Z. Zhu, D. Xu, and Z. Wang. Lightgaussian: Unbounded 3d gaussian compres- sion with 15x reduction and 200+ fps. arXiv preprint arXiv:2311.17245, 2023. 2

    2023 arXiv

  2. [10]

    Z. Fan, W. Cong, K. Wen, K. Wang, J. Zhang, X. Ding, D. Xu, B. Ivanovic, M. Pavone, G. Pavlakos, Z. Wang, and Y . Wang. Instantsplat: Sparse-view sfm-free gaussian splatting in seconds. ArXiv, 2024. 2

    2024

  3. [11]

    Girish, K

    S. Girish, K. Gupta, and A. Shrivastava. Eagles: Efficient accelerated 3d gaussians with lightweight encodings. arXiv preprint arXiv:2312.04564, 2023. 6

    2023 arXiv

  4. [12]

    Relaxing accu- rate initialization constraint for 3d gaussian splatting

    Jaewoo Jung, Jisang Han, Honggyu An, Jiwon Kang, Seonghoon Park, and Seungryong Kim. Relaxing accu- rate initialization constraint for 3d gaussian splatting. arXiv preprint arXiv:2403.09413, 2024. 2

    2024 arXiv

  5. [13]

    3d gaussian splatting for real-time ra- diance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time ra- diance field rendering. ACM Transactions on Graphics (SIG- GRAPH Conference Proceedings), 42(4), 2023. 1, 3

    2023

  6. [14]

    Deblurring 3d gaussian splatting, 2024

    Byeonghyeon Lee, Howoong Lee, Xiangyu Sun, Usman Ali, and Eunbyung Park. Deblurring 3d gaussian splatting, 2024. 4

    2024

  7. [15]

    J. Li, Y . Li, C. Sun, C. Wang, and J. Xiang. Spec-nerf: Multi- spectral neural radiance fields. ArXiv, 2023. 2

    2023

  8. [16]

    R. Li, J. Liu, G. Liu, S. Zhang, B. Zeng, and S. Liu. Spec- tralnerf: Physically based spectral rendering with neural ra- diance field. ArXiv, 2023. 2

    2023

  9. [17]

    3d plant modelling via hyperspectral imaging

    Jie Liang, Ali Zia, Jun Zhou, and Xavier Sirault. 3d plant modelling via hyperspectral imaging. In 2013 IEEE Inter- national Conference on Computer Vision Workshops, pages 172–177, 2013. 2

    2013

  10. [18]

    W. Liu, T. Guan, B. Zhu, L. Ju, Z. Song, D. Li, Y . Wang, and W. Yang. Efficientgs: Streamlining gaussian splatting for large-scale high-resolution scene representation. ArXiv,

  11. [19]

    T. Lu, M. Yu, L. Xu, Y . Xiangli, L. Wang, D. Lin, and B. Dai. Scaffold-gs: Structured 3d gaussians for view-adaptive rendering. ArXiv, 2023. 2, 4

    2023

  12. [21]

    Malarz, W

    D. Malarz, W. Smolak, J. Tabor, S. Tadeja, and P. Spurek. Gaussian splatting with nerf-based color and opacity. ArXiv,

  13. [22]

    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 syn- thesis. In ECCV, 2020. 1, 6

    2020

  14. [23]

    Mildenhall, P

    B. Mildenhall, P. Hedman, R. Martin-Brualla, P.P. Srini- vasan, and J.T. Barron. Nerf in the dark: High dynamic range view synthesis from noisy raw images. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 2

    2022

  15. [24]

    M ¨uller, A

    T. M ¨uller, A. Evans, C. Schied, and A. Keller. Instant neu- ral graphics primitives with a multiresolution hash encoding. ACM Transactions on Graphics (TOG), 41(4):102:1–102:15,

  16. [25]

    Niedermayr, J

    S. Niedermayr, J. Stumpfegger, and R. Westermann. Com- pressed 3d gaussian splatting for accelerated novel view syn- thesis. arXiv preprint arXiv:2401.02436, 2023. 6

    2023 arXiv

  17. [26]

    Tancik, E

    M. Tancik, E. Weber, E. Ng, R. Li, B. Yi, J. Kerr, T. Wang, A. Kristoffersen, J. Austin, K. Salahi, A. Ahuja, D. McAl- lister, and A. Kanazawa. Nerfstudio: A modular frame- work for neural radiance field development. arXiv preprint arXiv:2302.04264, 2023. 6

    2023 arXiv

  18. [27]

    Raspectloc: Ra- man spectroscopy-dependent robot localisation

    Christopher Thirgood, Oscar Mendez, Erin Chao Ling, Jon Storey, and Simon Hadfield. Raspectloc: Ra- man spectroscopy-dependent robot localisation. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5296–5303, 2023. 2

    2023

  19. [28]

    S. Wang, V . Leroy, Y . Cabon, B. Chidlovskii, and J. Revaud. DUSt3R: Geometric 3D Vision Made Easy. arXiv preprint arXiv:2312.14132, 2023. 2 9

    2023 arXiv

  20. [29]

    Z. Yang, X. Gao, Y . Sun, Y . Huang, X. Lyu, W. Zhou, S. Jiao, X. Qi, and X. Jin. Spec-Gaussian: Anisotropic View- Dependent Appearance for 3D Gaussian Splatting. ArXiv,

  21. [30]

    Mip-splatting: Alias-free 3d gaussian splat- ting

    Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, and Andreas Geiger. Mip-splatting: Alias-free 3d gaussian splat- ting. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 19447– 19456, 2024. 2

    2024

  22. [31]

    A. Zia, J. Liang, J. Zhou, and Y . Gao. 3d reconstruction from hyperspectral images. In 2015 IEEE Winter Conference on Applications of Computer Vision (WACV) , pages 318–325,

    2015

  23. [32]

    3d recon- struction from hyperspectral images

    Ali Zia, Jie Liang, Jun Zhou, and Yongsheng Gao. 3d recon- struction from hyperspectral images. In 2015 IEEE Winter Conference on Applications of Computer Vision, pages 318– 325, 2015. 2 10 HyperGS: Hyperspectral 3D Gaussian Splatting Supplementary Material In this supplementa...

    2015

  24. [33]

    Our method consis- tently enhances scene modeling performance on almost ev- ery combination of metric and scenarios

    Per-scene quantitative results Tables 4 and 5 break down the results of the SOP and Bayspec datasets results of Tables 2 and 1 from the main paper, into metrics for each scene. Our method consis- tently enhances scene modeling performance on almost ev- ery combination of metri...

  25. [34]

    turntable style

    Simulated Scannet dataset Since no room-scale multi-view hyperspectral dataset cur- rently exists, we propose simulating such a dataset using the ScanNetV2 dataset. To enhance the diversity and spec- tral richness, we utilize longer channel depths by incorpo- rating downsample...

  26. [35]

    We found that refining the technique on this dataset yielded greater results in the SOP dataset that we tested on as part of the main paper

    Additional ablation studies In this section we aim to provide greater ablation studies on the bayspec dataset for HyperGS. We found that refining the technique on this dataset yielded greater results in the SOP dataset that we tested on as part of the main paper. 9.1. Pruning ...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.