Pith. sign in

REVIEW 6 major objections 6 minor 27 references

WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting

T0 review · 6 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 Per-vertex spherical-harmonic reconstruction with a mesh-local probe association delivers static global illumination at roughly 5 percent of the memory of industry lightmap methods, while cutting fragment-shader sampling.

desk verdict A well-engineered mobile GI baking paper with a genuinely new per-mesh probe-association idea, but the generalization of that fixed association is under-analyzed and the metric slightly favors the method. read the letter →

arxiv 2506.01288 v1 pith:PVQGAEIV submitted 2025-06-02 cs.GR

Junke Zhu , Zehan Wu , Qixing Zhang , Cheng Liao , Zhangjin Huang This is my paper
classification cs.GR
keywords staticglobalilluminationsphericalharmonicsinverseprobedistributionlightbakinglow-endplatformslightmapfragmentshadersamplinglevelofdetail
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 proposes a static global illumination baking pipeline aimed at low-end real-time platforms. It claims that reconstructing light per vertex from a small set of spherical-harmonic probes, with each mesh assigned an offline-optimized probe association in its own local space, produces lighting quality competitive with mainstream industry lightmap methods while using only about 5 percent of their memory. The method also reduces fragment-shader texture sampling and removes the need for UV mapping. If the claim holds, mobile and other constrained devices could run convincing baked global illumination in forward rendering without an extra light pass.

What carries the argument

The load-bearing object is the per-mesh association $A$: for each vertex, a short list of probe indices and normalized weights that defines the vertex's spherical harmonics as a linear combination of probe spherical harmonics. It is initialized by K-medoids clustering on the mesh surface with a visibility-aware distance (Euclidean for mutually visible points, pathfinding distance for occluded ones), then refined by minimizing the lighting loss over rotated views of the mesh in a standard scene. This association removes scene-space probe placement and lets all instances of one mesh share a single lightweight embedding; the spherical-harmonic coefficients themselves are then solved from the normal equations of a linear least-squares problem, Eq. (9), with the regularization controlled by $\lambda$.

What would settle it

Bake one mesh with the fixed association under adversarial real scenes—strong directional sources, large occluders, and light distributions absent from the rotation-based training—then compare its mRMSE and visible light leakage against the same mesh with probes optimized in that actual scene; a large quality gap would falsify the generalization claim.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that pixel-level probe interpolation is unnecessary for static objects: a vertex-based illumination model, where each vertex's spherical-harmonic vector is a weighted combination of probes and fragments interpolate those vectors barycentrically, can carry the full reconstruction. The probes are fit by minimizing a hemisphere-weighted lighting error $\mathbf{E}_{\mathrm{light}}$ plus a gradient-regularization term, which reduces baking to a linear least-squares problem in the probe coefficients. The other half is the inverse probe distribution: K-medoids clustering with a visibility-aware distance initializes the vertex–probe association $A$, and gradient descent in a synthetic standard scene refines it, so the association embeds into the mesh and every instance reuses it. The paper reports lower multi-directional RMSE than Directional Lightmap and Volumetric Lightmap on most test meshes, with fragment texture samples per texel dropping to 3.08 from 12.2 (VLM) and 5.14 (Lightmap).

Load-bearing premise

The load-bearing premise is that one per-mesh vertex–probe association, tuned once in a synthetic standard scene and reused by every instance, keeps its quality in arbitrary real scenes; if the optimal association depends on the actual arrangement of lights and occluders, the fixed weights will leak light or blur shading in layouts unlike the training set.

Editorial extensions

If this is right

  • Static scene lighting can be baked and rendered at roughly 5 percent of the memory of Directional Lightmap or Volumetric Lightmap configurations at comparable or better measured mRMSE.
  • Fragment-shader texture sampling drops to about 3.08 samples per texel, versus 12.2 for Volumetric Lightmaps and 5.14 for Directional Lightmaps, which directly relieves the bottleneck on tile-based deferred GPUs.
  • Artists can skip UV mapping for static geometry, since illumination is stored at vertices and in a single scene-wide probemap, eliminating UV-gap waste and draw-call overhead.
  • Level-of-detail support is built in by switching between second-order and third-order spherical harmonics, taking one or two texture samples depending on distance.
  • Time-of-day updates reduce to interpolating among a small set of probemaps, with a day-night cycle demonstrated at 4 MB total for eight time points.

Reading between the lines

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

  • The paper does not quantify how the fixed association $A$ degrades when a scene's lighting layout differs strongly from the synthetic standard scene; a sweep over lighting configurations, with per-scene optimized probes as the reference, would reveal the true generalization margin.
  • A natural testable extension is dynamic objects: interpolating spherical harmonics from nearby probes, which the paper sketches as viable, could be validated against path-traced ground truth for moving characters.
  • The paper itself notes in Section 5.4 that spherical harmonics and the vertex-probe structure cannot represent sharp shadows and that LOD transitions may pop; these are acknowledged boundaries that bound the 'competitive lighting' claim to low-frequency, static lighting.
  • The probe-count ceiling of 256 raises a packing question the paper leaves open: how many distinct meshes can share one scene probemap before texture resolution or index precision becomes the limiting factor.
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

6 major / 6 minor

Summary. This paper presents WishGI, a static global illumination baking pipeline for low-end platforms that reconstructs per-vertex hemispherical illumination as a linear combination of a small number of scene-level spherical-harmonics probes. The probe coefficients are obtained by solving a regularized least-squares fit to ray-traced radiance over the effective hemisphere, and the vertex–probe association is optimized offline in the mesh's local space by rotating the mesh in a synthetic standard scene. The method claims roughly 5% of the memory of mainstream lightmap/volumetric lightmap approaches, lower fragment-shader texture sampling, and competitive visual quality, with experiments on three Unreal scenes and several single meshes.

Significance. If the claims hold, the work is a practical contribution: the local-space inverse probe association is a novel idea that could reduce probe redundancy, and the reported fragment-sampling reduction (3.08 vs. 12.2 samples in Table 3) is a plausible path to low-end support. The optimization is differentiable and the core formulas are standard least squares, which lends credibility to the method. However, the current evidence leaves the generalization of the fixed association insufficiently quantified, which is central to the claim of consistent quality.

major comments (6)
  1. [Section 3.4 / Eq. (10)] The vertex–probe association A is optimized in a single 'standard scene' with unspecified complexity, lighting, and rotation set, and no sensitivity analysis is reported. Because A is then fixed for all scenes and instances, the paper's central claim of consistent lighting quality depends on this generalization. Specify the standard scene, the number of scenarios N_sc, and the rotation set, and report mRMSE or visual results across a range of lighting environments (e.g., directional, point, area lights, varying occlusion) and scene structures.
  2. [Section 5.1 / Eq. (13) vs. Section 3.3 / Eq. (4)] The evaluation metric mRMSE is essentially the same cosine-weighted hemispherical difference that the baking loss minimizes (Eq. 4). The method is therefore being evaluated on its own objective, which biases the quantitative comparison in Table 2. Please add an independent metric, such as per-pixel PSNR/SSIM of rendered images against a path-traced reference, or demonstrate that the weighting choice does not change the ranking of methods.
  3. [Section 5.1 / Ground truth] The numerical ground truth is the third-order SH projection of ray-traced radiance, and the method itself reconstructs SH2/SH3 coefficients. This shared representation can favor the method. Please justify the choice of SH order for the ground truth, and either report the error against the original ray-traced radiance or use a representation-agnostic metric to confirm the quantitative results.
  4. [Abstract / Section 5.2 / Section 5.4] The headline 'approximately 5% of the memory' is not substantiated. Table 2 reports per-object probe memory only; the scene-level Time-of-Day example in Section 5.4 shows 4MB for eight probemaps versus 18MB for a single lightmap, which is about 22% of the memory for one time point. Table 3 reports bandwidth in MB/frame, not storage. Provide a scene-level storage comparison that supports the 5% figure.
  5. [Eq. (9)] The derivation of the normal equation (9) is deferred to a supplementary that is not available to the reader. Since this is the core of the baking pipeline and the notation (e.g., the products involving w, T(Y), B, W) is non-trivial, please include the full derivation in the main text or in an accessible supplementary document.
  6. [Section 5.4] The limitation paragraph concedes that for meshes that are 'highly complex' or 'do not require reuse', targeted optimization based on the actual lighting environment is preferable. This directly qualifies the claim that the fixed association A works consistently for all instances. Please characterize these failure cases quantitatively (e.g., which of the tested meshes required targeted optimization) and incorporate this into the central claim.
minor comments (6)
  1. [Section 3.4] K-Medoids is attributed to Lloyd [1982], which is the reference for k-means; please use a k-medoids reference or clarify the clustering variant.
  2. [Section 3.2] The phrase 'we use twelve 10-bits' should be 'twelve 10-bit values' and similarly for 'fifteen 8-bits'.
  3. [Table 2] The column header 'Mesh | Mem.' appears to combine the mesh name and memory; reformat for readability.
  4. [Section 3.2 / Figure 4] The terms 'pixel' for the two 16-byte blocks in Figure 4 are confusing; use 'texel' or 'data block'.
  5. [Section 2] There is an unresolved placeholder '[Knodt et al. 2023; ?]'; fill in the missing reference.
  6. [Section 3.4] The sentence 'Each vertex needs to be assigned the top n probes with the highest weights as geometric priors to avoid visual discontinuities in Figure 7(a) when the mesh triangles are dense' is unclear; specify what 'top n' is and how it relates to the two-probe-per-vertex setup.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SH coefficients and vertex-probe associations are empirical fits compared against independent baselines; the metric/loss similarity is an evaluation choice, not a definitional reduction.

full rationale

The paper's derivation chain is a standard baking pipeline. Eqn. (8) solves for probe SH coefficients by least-squares fitting to ray-traced illumination samples, with the normal equations given in Eqn. (9); the target is actual illumination f, not a quantity defined in terms of the fitted SH. Eqn. (10) optimizes the vertex-probe association A against an E_light objective over synthetic rotated scenarios, and this association is then fixed and evaluated in separate Unreal scenes. Neither step defines its output as its input: fhat depends on A and SH, but f is external data. The memory claim follows from the storage format (two 8-bit indices/weights per vertex plus a compact probemap), not from a circular equation. The mRMSE metric in Eqn. (13) shares cosine hemisphere weighting with the baking loss in Eqn. (4), and the numerical ground truth is itself a third-order SH fit; these are evaluation choices that may favor SH-based reconstructions, but they do not make any derived result equivalent to its inputs by construction. The comparisons to Directional Lightmap and Volumetric Lightmap are made against the same source data and external implementations. No load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation are present. The generalization risk of the fixed association A is an empirical validation concern, not a circularity.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The method's load-bearing assumptions are modest and mostly standard for probe-based GI: SH sufficiency, a cosine weighting for effective directions, and a heuristic standard scene for association training. No new physical entities are introduced. The largest unverified premise is the generalization of the per-mesh association to unseen scenes, which is an assumption about the training distribution rather than a free numerical parameter.

free parameters (8)
  • number_of_probes_K = 256 max, typically 20-50
    Chosen by hand; Table 2 uses 20 and 30 probes per mesh; Section 4 states 256 probes are sufficient for most objects.
  • sampling_density = 100 points per m^2
    Set empirically in Section 4; Figure 8 shows quality saturates beyond this value, but the number itself is a fixed choice.
  • regularization_weight_lambda = 0.1
    Set in Section 4; no sensitivity analysis beyond the ablation in Figure 9.
  • per_vertex_probe_count = 2
    Each sampling point is associated with two probes; chosen to balance smoothness and memory, as described in Section 4.
  • SH_orders = second order for LOD1, third order for LOD0
    Encoding choice in Section 3.2; determines quality and memory.
  • sampling_directions_bake = 960
    Number of spherical directions for the baking loss, set in Section 4.
  • sampling_directions_probe = 120
    Directions used in probe distribution optimization, with angles less than 30 degrees from the normal, set in Section 4.
  • optimizer_iterations = 400
    Adam for 400 iterations; Figure 10 shows 400 is enough with the geometric prior, but this is an empirical choice.
assumptions (5)
  • domain assumption Illumination f is an integrable function from S to R^3
    Defined in Section 3.1; standard assumption for rendered radiance fields, but it presupposes the signal is representable by the chosen basis.
  • domain assumption Third-order spherical harmonics suffice for static GI quality
    Used throughout; the ground truth itself is a third-order SH fit, and encoding uses orders 2 and 3, so the method cannot represent high-frequency content beyond SH-3 (Section 5.4 admits this).
  • ad hoc to paper Weighting w(d)=max(0, cos(d,n)) correctly captures effective illumination directions
    Introduced in Eqn. (4); favors normal-aligned directions, which is reasonable for normal-map shading but is a design choice not derived from a physical principle.
  • standard math The least-squares problem in Eqn. (9) is well-posed and its normal equations give the global minimum
    The objective is quadratic in SH, so solving for zero gradient gives the minimum; the detailed derivation is in the supplementary.
  • ad hoc to paper The synthetic standard scene and rotation set in Section 3.4 cover the relevant lighting conditions for all real scenes
    The generalization of the association A from a few synthetic scenarios to arbitrary scenes is assumed, not demonstrated by sensitivity analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting." pith.science (2026). https://pith.science/paper/PVQGAEIV

@misc{pith2026250601288,
  author       = {Pith},
  title        = {Pith review of: WishGI: Lightweight Static Global Illumination Baking via Spherical Harmonics Fitting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PVQGAEIV}},
  note         = {Machine review of arXiv:2506.01288}
}
read the original abstract

Global illumination combines direct and indirect lighting to create realistic lighting effects, bringing virtual scenes closer to reality. Static global illumination is a crucial component of virtual scene rendering, leveraging precomputation and baking techniques to significantly reduce runtime computational costs. Unfortunately, many existing works prioritize visual quality by relying on extensive texture storage and massive pixel-level texture sampling, leading to large performance overhead. In this paper, we introduce an illumination reconstruction method that effectively reduces sampling in fragment shader and avoids additional render passes, making it well-suited for low-end platforms. To achieve high-quality global illumination with reduced memory usage, we adopt a spherical harmonics fitting approach for baking effective illumination information and propose an inverse probe distribution method that generates unique probe associations for each mesh. This association, which can be generated offline in the local space, ensures consistent lighting quality across all instances of the same mesh. As a consequence, our method delivers highly competitive lighting effects while using only approximately 5% of the memory required by mainstream industry techniques.

Figures

Figures reproduced from arXiv: 2506.01288 by the authors.

Figure 1
Figure 1. We present a novel method for static global illumination baking tailored for low-end platforms. Our work significantly reduces memory usage and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Normal maps perturb the geometric normals in tangent space, and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. (a) The first pixel (16 bytes) stores the Multiplier and the coefficients [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (9 more)
Figure 5
Figure 5. Figure 5: We fit the values for effective directions of valid sample points, [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 8
Figure 8. Figure 8: Visual artifacts are noticeable at lower sampling densities. Beyond [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 7
Figure 7. Figure 7: (a) Each sampling point is associated only with the nearest probe, [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 9
Figure 9. Figure 9: The Impact of Gradient Regularization (Forge with 40 probes, 1.25 [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: The Impact of Geometric Prior (Statue with 50 probes, 1.56KB). [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: The probemap at a given moment is derived through interpolation [PITH_FULL_IMAGE:figures/full_fig_p008_11.png]
Figure 13
Figure 13. Figure 13: The results of multiple objects in a Cornell Box with the left and right walls as emissive light sources. Our work effectively reduces light leakage and [PITH_FULL_IMAGE:figures/full_fig_p009_13.png]
Figure 14
Figure 14. Figure 14: The results of single mesh. Each mesh was exposed to over three light sources. Our work was tested on various types of models with different [PITH_FULL_IMAGE:figures/full_fig_p009_14.png]
Figure 15
Figure 15. Figure 15: The results of entire scenes. We conducted tests in Unreal Engine’s Sun Temple, Summer HandPainted Environment, and Stylized Egypt. Compared to [PITH_FULL_IMAGE:figures/full_fig_p010_15.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 25 canonical work pages

  1. [3]

    In ACM SIGGRAPH 2008 Games. 1–22. Zhonggui Chen, Zhan Yuan, Yi-King Choi, Ligang Liu, and Wenping Wang

    work page 2008

  2. [6]

    ACM Trans

    Efficient Light Probes for Real-Time Global Illumination. ACM Trans. Graph. 41, 6, Article 202 (Nov. 2022), 14 pages. Peter E. Hart, Nils J. Nilsson, and Bertram Raphael

    work page 2022

  3. [11]

    ACM Trans

    Joint UV Optimization and Texture Baking. ACM Trans. Graph. 43, 1, Article 2 (Sept. 2023), 20 pages. https://doi.org/10.1145/3617683 Dimitar Lazarov

    doi:10.1145/3617683 2023

  4. [14]

    ACM Transactions on Graphics 42, 3 (2023), 1–15

    Spatiotemporally consistent hdr indoor lighting estimation. ACM Transactions on Graphics 42, 3 (2023), 1–15. Ligang Liu, Lei Zhang, Yin Xu, Craig Gotsman, and Steven J Gortler

    work page 2023

  5. [17]

    Journal of Computer Graphics Techniques 8, 2 (2019)

    Dynamic diffuse global illumination with ray-traced irradiance fields. Journal of Computer Graphics Techniques 8, 2 (2019). Air Silvennoinen Michal Iwanicki, Peter-pike Sloan and Peter Shirley

    work page 2019

  6. [19]

    The State of the Art in Interactive Global Illumination. Comput. Graph. Forum 31, 1 (Feb. 2012), 160–188. Carlos Rodriguez-Pardo, Javier Fabre, Elena Garces, and Jorge Lopez-Moreno

    work page 2012

  7. [20]

    Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations

    Scene Represen- tation Networks: Continuous 3D-Structure-Aware Neural Scene Representations. arXiv:1906.01618 [cs.CV] https://arxiv.org/abs/1906.01618 Peter-Pike Sloan

    work page Pith review arXiv 1906

  8. [22]

    ACM Transactions on Graphics (TOG) 24, 3 (2005), 1216–1224

    Local, deformable precomputed radiance transfer. ACM Transactions on Graphics (TOG) 24, 3 (2005), 1216–1224. Peter-Pike Sloan and Ari Silvennoinen

    work page 2005

Show all 27 references
  1. [26]

    In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    NeLF-Pro: Neural Light Field Probes for Multi-Scale Novel View Synthesis. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 19833–19843. Fangneng Zhan, Changgong Zhang, Yingchen Yu, Yuan Chang, Shijian Lu, Feiying Ma, and Xuansong Xie

    2024

  2. [35]

    ACM Trans

    3287–3295. ACM Trans. Graph., Vol. 44, No. 4, Article . Publication date: August 2025

    2025

  3. [1968]

    IEEE Transactions on Systems Science and Cybernetics 4, 2 (1968), 100–107

    A Formal Basis for the Heuristic Determination of Minimum Cost Paths. IEEE Transactions on Systems Science and Cybernetics 4, 2 (1968), 100–107. John T Hooker

    1968

  4. [1982]

    IEEE Transactions on Information Theory 28, 2 (1982), 129–137

    Least squares quantization in PCM. IEEE Transactions on Information Theory 28, 2 (1982), 129–137. Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai

    1982

  5. [1991]

    The JPEG still picture compression standard. Commun. ACM 34, 4 (1991), 30–44. Youxin Xing, Gaole Pan, Xiang Chen, Ji Wu, Lu Wang, and Beibei Wang

    1991

  6. [2002]

    ACM Trans

    Precomputed radiance transfer for real-time rendering in dynamic, low-frequency lighting environments. ACM Trans. Graph. 21, 3 (July 2002), 527–536. https://doi.org/10.1145/566654.566612 Peter-Pike Sloan, Ben Luna, and John Snyder

    2002

  7. [2005]

    In ACM SIGGRAPH 2005 Sketches

    Optimized photon tracing using spherical harmonic light maps. In ACM SIGGRAPH 2005 Sketches. 53–es. Jonathan Granskog, Fabrice Rousselle, Marios Papas, and Jan Novák

    2005

  8. [2011]

    SIGGRAPH Advances in Real-Time Rendering in Games course (2011)

    Physically-based lighting in call of duty: Black ops. SIGGRAPH Advances in Real-Time Rendering in Games course (2011). Yue Li, Pablo Wiedemann, and Kenny Mitchell

    2011

  9. [2012]

    IEEE Transactions on Visualization and Computer Graphics 18, 10 (2012), 1784–1796

    Variational Blue Noise Sampling. IEEE Transactions on Visualization and Computer Graphics 18, 10 (2012), 1784–1796. Robert Cupisz

    2012

  10. [2013]

    In ACM SIGGRAPH 2013 Talks

    Lighting technology of the last of us. In ACM SIGGRAPH 2013 Talks. Michał Iwanicki and Peter-Pike Sloan. 2017a. Ambient Dice. InEurographics Symposium on Rendering. Michał Iwanicki and Peter-Pike Sloan. 2017b. Precomputed lighting in Call of Duty: Infinite Warfare. SIGGRAPH Ad...

    2017

  11. [2016]

    Michal Iwanicki

    Volumetric global illumination at Treyarch.Advances in Real-Time Rendering (2016). Michal Iwanicki

    2016

  12. [2017]

    ACM Transactions on Graphics (TOG) 36, 6 (2017), 216:1–216:15

    Seamless: seam erasure and seam-aware decoupling of shape from mesh resolution. ACM Transactions on Graphics (TOG) 36, 6 (2017), 216:1–216:15. S. Lloyd

    2017

  13. [2018]

    In SIGGRAPH Asia 2018 Technical Briefs (Tokyo, Japan) (SA ’18)

    Directional lightmap encoding insights. In SIGGRAPH Asia 2018 Technical Briefs (Tokyo, Japan) (SA ’18). Association for Computing Machinery, New York, NY, USA, Article 12, 3 pages. Shuran Song and Thomas Funkhouser

    2018

  14. [2019]

    Deep Precomputed Radiance Transfer for Deformable Objects. Proc. ACM Comput. Graph. Interact. Tech. 2, 1, Article 3 (June 2019), 16 pages. https://doi.org/10.1145/3320284 Zhengqin Li, Li Yu, Mikhail Okunev, Manmohan Chandraker, and Zhao Dong

    2019 doi

  15. [2020]

    ACM Trans

    Composi- tional neural scene representations for shading inference. ACM Trans. Graph. 39, 4, Article 135 (Aug. 2020), 13 pages. Jie Guo, Zijing Zong, Yadong Song, Xihao Fu, Chengzhi Tao, Yanwen Guo, and Ling-Qi Yan

    2020

  16. [2022]

    ACM Transactions on Graphics (TOG) 41, 5 (2022), 1–18

    Active exploration for neural global illumination of variable scenes. ACM Transactions on Graphics (TOG) 41, 5 (2022), 1–18. Epic Games. 2024a. Indirect Lighting Cache in Unreal Engine. https: //dev.epicgames.com/documentation/en-us/unreal-engine/indirect-lighting- cache-in-un...

    2022

  17. [2023]

    ACM Trans

    3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph. 42, 4 (2023), 139–1. Julian Knodt, Zherong Pan, Kui Wu, and Xifeng Gao

    2023

  18. [2024]

    SIGGRAPH 2024 Course

    Neural Light Grid. SIGGRAPH 2024 Course. Gilles Rainer, Adrien Bousseau, Tobias Ritschel, and George Drettakis

    2024

  19. [2025]

    https://doi.org/10.1007/s41095-023-0367-z Zinuo You, Geiger andreas, and Anpei Chen

    12 • Junke Zhu, Zehan Wu, Qixing Zhang, Cheng Liao, and Zhangjin Huang Media 10, 5 (2024), 923–936. https://doi.org/10.1007/s41095-023-0367-z Zinuo You, Geiger andreas, and Anpei Chen

    2024 doi

Pith tools

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