Towards Physically-Based Sky-Modeling

Ian J. Maquignaz

REVIEW 3 major objections 6 minor 52 references

Towards Physically-Based Sky-Modeling

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 AllSky, a DNN trained on physically captured HDR skies, generates environment maps that retain more of the sun's dynamic range than prior deep sky-models, using exposure-bracketed class-aware losses and a learned LDR-to-EDR booster.

desk verdict A promising set of HDR-aware losses and a learned LDR-to-EDR head, but the illumination claim is contradicted by the paper's own tables and the EV metric is nonstandard. read the letter →

arxiv 2412.11883 v1 pith:H7OOQEWZ submitted 2024-12-16 cs.CV eess.IV

Ian J. Maquignaz This is my paper · ORCID

classification cs.CVeess.IV

keywords sky-modelsuncloudshighdynamicrangeenvironmentmapsimage-basedlightingextendedHDRimagesynthesis

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 argues that previous deep-learning sky-models, despite improving visual quality, quietly lose the wide dynamic range of real outdoor light, so generated environment maps relight scenes with wrong tones and shadows. It proposes AllSky, a conditional generator trained on tens of thousands of physically captured HDR skies, paired with two mechanisms: losses that weight errors by exposure bracket and by region (sun, sky, cloud) and a learned inverse-tone-mapping network that expands a low-dynamic-range output back toward 14 stops. In comparisons, AllSky reports higher ratios of generated-to-ground-truth dynamic range and integrated illumination than a cloud-augmenting baseline and a text-driven HDR panorama model. The paper's deeper point is that dynamic range, not average pixel error, is what makes an environment map usable for image-based lighting.

What carries the argument

The load-bearing mechanism is a two-stage training setup. First, a hybrid tone-mapper, μ-lawLog2, compresses HDR intensities to a range near −1 to 1, preserving low-exposure cloud texture while aggressively compressing saturated solar regions. Second, the loss function is split by exposure bracket (Cascade loss) and by semantic class (Selective loss), so the small, ultra-bright sun region is not numerically drowned out by the large dim sky. Third, an ANN head called ldr2EDR (or latent2EDR, operating on the U-Net's latent space) learns the inverse tone-mapping from the network's compressed output back to linear EDR values. The paper evaluates with the EV ratio and the integrated illumination ratio rather than relying on L1, L2, or perceptual metrics alone.

What would settle it

Recompute the headline comparisons with EV defined as log2(max/min) and with integrated illumination measured after exposure equalization; if AllSky no longer beats DeepClouds and Text2Light on those recomputed ratios, the central claim that it retains more extended dynamic range fails.

Watch

Extended reading notes

Core claim

The central claim is that a DNN sky-model can learn weathered skies directly from physically captured HDR imagery and, with the right loss structure and a learned decompression head, retain a substantially larger fraction of the scene's extended dynamic range than prior DNN models. The paper shows that conventional L1 and L2 training in tonemapped space collapses the solar region's intensity, and that adding Selective and Cascade losses—which split the image into exposure bands and semantic classes—plus an ANN LDR2EDR booster recovers a larger share of ground-truth EV and integrated illumination. Reported results put AllSky's generated-to-ground-truth EV ratios near 0.85–1.12 and illumination ratios near 0.5, versus about 0.36 EV for the DeepClouds baseline and a Text2Light HDR overshoot of 2.42 EV and 3.55 times illumination, on the paper's own metrics.

Load-bearing premise

The load-bearing premise is that the paper's EV formula, log2(|I|max − |I|min), is a meaningful measure of the dynamic range that matters for image-based lighting, since all headline comparisons use it and the standard log-ratio definition would give different numbers.

Editorial extensions

If this is right

AllSky outputs, conditioned on a user-drawn sun and cloud label, retain a larger fraction of ground-truth EV and integrated illumination than DeepClouds and Text2Light, so rendered scenes should show tones and shadows closer to the physical capture.
Cascade and Selective losses can be dropped into other HDRI generation pipelines, since they only replace the loss function rather than the architecture.
The learned LDR2EDR head can extend the dynamic range of an already-trained U-Net backbone with negligible change to visual-quality metrics such as LPIPS and FID.
Dynamic range and integrated illumination should be reported alongside L1/L2 and perceptual scores when evaluating any sky-model, because conventional metrics are shown to be insensitive to the clipping that changes image-based lighting.
The parametric clear-sky sun aggregation used by DeepClouds is what gives it dynamic range, but the paper finds it produces a tiny, visually unappealing solar disc that can pierce clouds, a failure AllSky avoids by learning the sun from data.

Reading between the lines

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

If the dynamic-range comparisons are recomputed with the standard log-ratio definition of EV, log2(max/min), the gap between AllSky and its baselines may shrink, because the paper's formula log2(|I|max − |I|min) rewards intensity differences rather than contrast ratios.
The LDR2EDR head is color-agnostic and trained on a fixed dataset, so it could likely be transferred as a post-processor to any generator whose output is tonemapped LDR, though its behavior on non-sky content is untested.
A learned cloud segmentation, or labels derived from a weather model, would probably make the Selective losses more effective than the color-ratio threshold, since the paper itself notes that segmentation is inconsistent across seasons.
For image-based lighting, the practical test is not the EV ratio but rendered appearance; a direct comparison of shadows and specular highlights at matched exposures would settle whether the extra dynamic range actually changes the relit scene.

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.

Desk Editor's Note

read the letter

Colleague,

This paper has a genuinely useful idea buried under a misleading headline. The authors propose exposure-segmented cascade losses and class-selective losses, plus a learned LDR-to-EDR expansion head (ldr2EDR / latent2EDR) that lets a U-Net train in tonemapped space without losing the sun's brightness. That is a real contribution – I haven't seen those specific components in the cited DeepClouds, SkyGAN, LM-GAN, or Text2Light. The paper also does a good job demonstrating that standard visual metrics (LPIPS, FID) are insensitive to dynamic range truncation, which is an important point for the IBL community. The evaluation of Text2Light's duplicates and seam artifacts is a useful side result.

The problems are in the evidence. The abstract only claims "improved retention of EDR," but the introduction claims "greater retention of dynamic range and illumination." The illumination half is not supported by their own numbers. Table 3 gives AllSky integrated-illumination ratios of 0.36–0.53, while Table 1 shows DeepClouds at 0.59 and DeepClouds without clear-sky at 0.62. The tables may use different training sets and resolutions, but the claim is unqualified and the paper does not discuss this discrepancy. That is a load-bearing flaw.

The dynamic range metric is also trouble. They define EV = log2(|I|max − |I|min), which is not a ratio and does not measure the number of stops between max and min. Standard EV/DR is log2(max/min). As defined, a scene with a bright sun and near-black sky could have the same EV as a uniform scene if the absolute spread is the same. And since their cascade losses explicitly supervise per exposure bracket, the EV ratio they report is essentially what the loss is optimizing – some circularity is unavoidable. They give no validation that this metric tracks usable dynamic range for IBL.

There is also no code, no error bars, and the AllSky experiments are at 64x64 on 5,000 images, far from the scale of the DeepClouds reproduction. The paper honestly acknowledges cloud-texture loss and segmentation inconsistencies, but the metric and claim issues are more serious than those limitations.

Who is this for? People working on sky models, HDR generation, and image-based lighting. The losses and the ldr2EDR idea are worth knowing about, even if the evaluation is not convincing. I would not desk-reject it: it has a novel technique and a valid problem. A careful reviewer could help the authors reframe the claims, fix the metric, and run apples-to-apples comparisons. Needs major revision, but there is something here.

Recommendation: send to peer review, with a strong request for code, error bars, and a metric justification.

Referee Report

3 major / 6 minor

Summary. The paper proposes AllSky, a conditional generative sky-model trained on the Laval HDR Sky database, which takes user-provided sun-position and cloud-formation label maps and outputs HDR environment maps. The method introduces a custom μ-lawLog2 tonemapper, exposure-bracketed 'Cascade' losses, class-segmented selective losses, and ANN-based LDR-to-EDR (ldr2EDR and latent2EDR) heads. The central claim, stated in Section 1 and the abstract, is that AllSky produces environment maps with greater retention of dynamic range and illumination than prior DNN sky-models. The paper evaluates this through an EV ratio metric defined as log2(|I|max - |I|min), an integrated illumination ratio, and a set of visual metrics.

Significance. If the central claim were correctly demonstrated, the work would be a useful step toward controllable, physically-plausible HDR sky environment maps for image-based lighting. The paper also raises a legitimate concern that standard LDR-space losses are insensitive to the high dynamic range of solar regions. However, the significance is heavily undermined by the evaluation methodology: the custom EV metric is nonstandard and unvalidated, the illumination-retention claim is contradicted by the paper's own tables, and the training objective is constructed so that the headline metric is directly optimized, creating a circularity concern. The paper does not ship code or pre-trained models, and visual results are acknowledged to lose cloud texture, so the practical contribution is currently limited.

major comments (3)

[Section 3.1 (also Introduction)] The dynamic-range metric is defined as EV = log2(|I|max - |I|min), i.e., the logarithm of an intensity difference, rather than the conventional definition EV = log2(max/min). This definition is scale-dependent and does not measure the number of stops between the dimmest and brightest parts of the image in any standard sense. All claims about 'Extended Dynamic Range' retention in Tables 2 and 3 are computed with this metric. No validation is provided that this quantity tracks usable dynamic range for image-based lighting, and no comparison to standard metrics (e.g., the ratio of the 99th to 1st percentile radiance, or the max/min ratio) is given. If the formula is unrepresentative, the headline quantitative results collapse. This is load-bearing and needs to be fixed or replaced.
[Section 4.3.4, Table 3 vs. Section 4.3.1, Table 1] The introduction claims that AllSky demonstrates 'greater retention of dynamic range and illumination' (Section 1, final paragraph). The illumination part is refuted by the paper's own measurements. Using the same integrated illumination ratio defined in Section 4.1.2 (Eq. 6), the best AllSky configuration in Table 3 (ldr2EDR-9 HDR) reaches an illumination ratio of 0.53, while Table 1 reports DeepClouds at 0.59 and DeepClouds w/o Clear-Sky at 0.62. Even ignoring the outlier DeepClouds w/ Sun (3e31, caused by the sun pass-through mechanism), AllSky does not beat the baseline on illumination. Since the abstract and introduction promise both dynamic-range and illumination retention, this internal contradiction materially undermines the central contribution. The authors should either revise the claim or provide a fair comparison that reconciles these numbers.
[Section 3.1.1, Eqs. (2)-(4) and Section 4.2.5] The Cascade losses explicitly segment the image by exposure brackets, with masks M = 2^{i-1} ≤ I ≤ 2^i and re-exposure by 2^{-i}, and penalize errors per bracket. The headline evaluation metric, EV ratio, is computed from the span of exposure values in the output. Training with these losses directly optimizes the quantity being measured: a model that matches the per-bracket intensity statistics will trivially improve the max-minus-min span. The reported EV gains therefore do not independently validate dynamic-range retention; they may simply reflect the training objective. To make the evaluation non-circular, the paper should either use an independent dynamic-range metric (e.g., percentile-based stops or a downstream IBL validation) or explicitly show that the metric improves on test data in a way that is not a direct artifact of the loss construction.

minor comments (6)

[Abstract and Section 1] The abstract promises 'improved retention of the Extended Dynamic Range (EDR) of the sky', while Section 1 promises 'greater retention of dynamic range and illumination'. The discrepancy matters because the illumination claim is not supported by the tables; the authors should align these statements.
[Section 4.1.3] The paper states that the cloud segmentation via color-ratio thresholding is 'robust but variable under different lighting and seasonality' and later acknowledges inconsistency (Section 5). Given that the segmentation directly supervises the class-aware losses, this limitation should be discussed more thoroughly, e.g., with an estimate of segmentation error on the test set.
[Section 5] The discussion concedes that 'cloud textures are essentially lost for all DNN sky-models', which is visually evident in Figures 8, 10 and 11. This is a significant caveat to any claim of 'photorealistic' output and should be stated more prominently, not only as a closing limitation.
[Table 3] The table reports FID for AllSky but Table 1 does not report FID for DeepClouds, making the visual-quality comparison incomplete. Also, several AllSky configurations have FID values above 100 (latent2EDR-9 HDR), which are much worse than typical trained generative models; this deserves commentary.
[Equations (2)-(4)] The exposure bracket masks are defined as M = 2^{i-1} ≤ I ≤ 2^i. For i=0 this includes negative intensities if I can be negative; the text says exposure is floored such that 2^{-1}=0, but the notation is not self-explanatory. Clarify the treatment of negative or zero intensities.
[Section 4.2.4] The Text2Light evaluation uses only 80 image pairs and matches them via ORB keypoints plus a scale-invariant loss. The paper should state the matching accuracy, since misalignment could affect the reported L1 and EV values. It would also help to report the number of unique prompts versus duplicated generated images (the text mentions 63 duplicates but does not give a clear breakdown).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the cascade losses target but do not define the EV metric, the held-out HDRDB evaluation is a standard split, and the only self-citation is background.

full rationale

The paper's central claim is that AllSky, trained with Selective and Cascade losses plus an ANN LDR2HDR head, improves retention of dynamic range and illumination. Walking the derivation chain, no step reduces to its own input by construction. The Cascade losses (Eqs. 2-5) segment the ground-truth HDRI into exposure brackets and penalize per-bracket reconstruction error; the headline EV metric (Section 3.1) is log2(|I|max - |I|min). These are not the same object: minimizing per-bracket L1/LPIPS does not by construction set the global max-min EV ratio, and the losses are not fitted parameters renamed as predictions. This is normal loss/metric alignment, not circularity. The ldr2EDR and latent2EDR heads are learned on HDRDB and evaluated on held-out dates from the same dataset, which is a conventional train/test split; the paper does not claim external generalization for AllSky beyond HDRDB, so this is not a fitted-input-called-prediction pattern. The only self-citation is reference [40] (LM-GAN), used in the background section to describe prior work; it is not load-bearing for AllSky's method, and no uniqueness theorem is imported from the authors. The cloud segmentation relies on Dev et al. [7], an external source. The paper's own tables do show that AllSky's best illumination ratio (0.53 in Table 3) is below DeepClouds values (0.59 and 0.62 in Table 1), which contradicts the abstract's 'retention of illumination' claim, but that is an internal-consistency and correctness concern, not a circularity concern under the stated rules. Because no equation reduces to itself and no fitted parameter is masquerading as a prediction, the circularity score is 0.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical and depends on the choice of metric, loss weights, and dataset assumptions. No new physical entities are introduced. The free parameters listed are those the paper sets by hand or leaves unspecified; all of them affect the reported dynamic-range results.

free parameters (7)

mu in mu-lawLog2 tonemapper = not reported
Eq. (1) defines the hybrid mu-lawLog2 operator, but the paper does not state the value of mu used in training and evaluation.
alpha-sun loss weight = 0.1
Used in LDR and HDR selective losses; chosen by hand to balance the small sun region.
alpha-skydome loss weight = 10
Used in HDR selective L1 loss for border and skydome; chosen by hand.
number of exposure brackets = 4
Global Cascade LPIPS uses 4 exposures; this is a design choice determining the loss sensitivity.
number of selective bands = 15
Selective Cascade uses 15 bands on the solar region; a free choice.
solar region mask size = 5 degrees
Extends the 0.5 degree solar disc to a 5 degree masked region; affects how much sun signal the losses target.
cloud segmentation threshold = not reported
Threshold on (B-R)/(B+R) from Dev et al.; the paper says segmentation is robust but variable, without giving the exact threshold.

assumptions (5)

domain assumption HDRDB images are accurate, calibrated physical captures of the sky.
The entire training and evaluation rest on HDRDB as ground truth; any calibration error biases all metrics.
standard math The tonemapping operators considered are bijections and thus losslessly invertible.
Stated in Section 3.1; needed for the LDR2HDR mapping to be recoverable.
domain assumption Ephemeris calculations give the correct solar position for each HDRDB capture.
Used to create solar masks in Section 4.1.3; if solar position is wrong, labels are wrong.
domain assumption The color-ratio cloud segmentation (B-R)/(B+R) yields usable cloud labels.
Section 4.1.3; the authors note it is robust but variable across seasons, which could add label noise.
ad hoc to paper A U-Net with 64x64 output can represent the distribution of outdoor skies.
The main experiments use 64x64 resolution; the paper does not demonstrate scalability to production resolutions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Towards Physically-Based Sky-Modeling." pith.science (2026). https://pith.science/paper/H7OOQEWZ

@misc{pith2026241211883,
  author       = {Pith},
  title        = {Pith review of: Towards Physically-Based Sky-Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H7OOQEWZ}},
  note         = {Machine review of arXiv:2412.11883}
}

read the original abstract

Accurate environment maps are a key component in rendering photorealistic outdoor scenes with coherent illumination. They enable captivating visual arts, immersive virtual reality and a wide range of engineering and scientific applications. Recent works have extended sky-models to be more comprehensive and inclusive of cloud formations but existing approaches fall short in faithfully recreating key-characteristics in physically captured HDRI. As we demonstrate, environment maps produced by sky-models do not relight scenes with the same tones, shadows, and illumination coherence as physically captured HDR imagery. Though the visual quality of DNN-generated LDR and HDR imagery has greatly progressed in recent years, we demonstrate this progress to be tangential to sky-modelling. Due to the Extended Dynamic Range (EDR) of 14EV required for outdoor environment maps inclusive of the sun, sky-modelling extends beyond the conventional paradigm of High Dynamic Range Imagery (HDRI). In this work, we propose an all-weather sky-model, learning weathered-skies directly from physically captured HDR imagery. Per user-controlled positioning of the sun and cloud formations, our model (AllSky) allows for emulation of physically captured environment maps with improved retention of the Extended Dynamic Range (EDR) of the sky.

Figures

Figures reproduced from arXiv: 2412.11883 by the authors.

**Figure 2.** Impact of dynamic range on scene illumination. Each column illus [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Tone-Mapping Operators In this regard, we propose a hybrid 𝜇-lawLog2 tonemapper as defined in eq. (1) and illustrate in fig. 3a. We propose this tonemapper to better preserve texture-rich clouds and skydome components (primarily 𝐸𝑉 ⪅ 1) and aggressively compress texture-poor high intensity regions which are primarily saturated in LDR imagery. 𝜇-lawLog2 (𝐼) = log2 log𝑒 (1 + 𝜇𝐼) log𝑒 (1.0 + 𝜇) + 1 (1) 3.1.1 Losse… view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: Database sample, June 7th, 2016 at 1:54PM [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 5.** Figure 5: AllSky Given the current paradigm, we theorize that DNNs trained in compressed-space are ignorant of the non-linear relationship between LDR and EDR space, and thus learn a depressed-LDR colorspace due to a presumption of constant error importance. To correct this, …

**Figure 7.** Figure 7: Text2Light: ‘purple petaled flower’. Arrow indicates latlong environment map seam. applicable, use our 𝜇-lawLog2 tone-mapping operator. As demonstrated in table 3, including our HDR-losses improves illumination and significantly improves dynamic range. For ldr2EDR an…

**Figure 8.** Figure 8: Images from modular training of latent2EDR [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

**Figure 9.** Figure 9: HDRDB labels paired to fig. 8 [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]

**Figure 10.** Figure 10: Text2Light: Visual matches for prompt ‘purple petaled flower’ [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

**Figure 11.** Figure 11: HDRI Generated by DeepClouds Baselines [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 39 canonical work pages

[1]

Khalid Alshaibani and Danny Li. 2021. Sky type classification for the ISO/CIE Stan- dard General Skies: a proposal for a new approach. International Journal of Low- Carbon Technologies 16, 3 (03 2021), 921–926. https://doi.org/10.1093/ijlct/ctab020 arXiv:https://academic.oup.com/ijlct/article-pdf/16/3/921/40490442/ctab020.pdf 8 • Ian J. Maquignaz

work page doi:10.1093/ijlct/ctab020 2021
[2]

Antoine Bouthors, Fabrice Neyret, Nelson Max, Eric Bruneton, and Cyril Crassin

work page
[3]

Eric Bruneton. 2017. A Qualitative and Quantitative Evaluation of 8 Clear Sky Models. IEEE Transactions on Visualization and Computer Graphics 23, 12 (2017), 2641–2655

work page 2017
[4]

Eric Bruneton and Fabrice Neyret. 2008. Precomputed Atmospheric Scattering. Computer Graphics Forum 27, 4 (2008), 1079–1086

work page 2008
[5]

Zhaoxi Chen, Guangcong Wang, and Ziwei Liu. 2022. Text2Light: Zero-Shot Text-Driven HDR Panorama Generation. ACM Transactions on Graphics (TOG) 41, 6, Article 195 (2022), 16 pages

work page 2022
[6]

Paul Debevec. 1998. Rendering synthetic objects into real scenes: Bridging tra- ditional and image-based graphics with global illumination and high dynamic range photography. In Proc. SIGGRAPH 98. 189–198

work page 1998
[7]

S. Dev, F. M. Savoy, Y. H. Lee, and S. Winkler. 2018. High-dynamic-range imaging for cloud segmentation. Atmospheric Measurement Techniques 11, 4 (2018), 2041–

work page 2018
[8]

David Eigen, Christian Puhrsch, and Rob Fergus. 2014. Depth Map Prediction from a Single Image using a Multi-Scale Deep Network. arXiv:1406.2283 [cs.CV]

arXiv 2014

Show all 52 references

[9]

Oskar Elek and Petr Kmoch. 2010. Real-time spectral scattering in large-scale nat- ural participating media. In Proceedings of the 26th Spring Conference on Computer Graphics. 77–84

2010
[10]

C. Emde, R. Buras-Schnell, A. Kylling, B. Mayer, J. Gasteiger, U. Hamann, J. Kylling, B. Richter, C. Pause, T. Dowling, and L. Bugliaro. 2016. The libRadtran software package for radiative transfer calculations (version 2.0.1). Geoscientific Model Development 9, 5 (2016), 1647...

2016 doi
[11]

David Griffiths, Tobias Ritschel, and Julien Philip. 2022. OutCast: Out- door Single-image Relighting with Cast Shadows. Computer Graph- ics Forum 41, 2 (2022), 179–193. https://doi.org/10.1111/cgf.14467 arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1111/cgf.14467

2022 doi
[12]

Jörg Haber, Marcus Magnor, and Hans-Peter Seidel. 2005. Physically-based simulation of twilight phenomena. ACM Transactions on Graphics (TOG) 24, 4 (2005), 1353–1373

2005
[13]

Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. 2017. GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium. InAdvances in Neural Information Processing Systems, I. Guyon, U. Von Luxburg, S. Bengio, H. Wall...

2017
[14]

Stephen Hill, Stephen McAuley, Laurent Belcour, Will Earl, Niklas Harrysson, Sébastien Hillaire, Naty Hoffman, Lee Kerley, Jasmin Patry, Rob Pieké, Igor Skliar, Jonathan Stone, Pascal Barla, Mégane Bati, and Iliyan Georgiev. 2020. Physically Based Shading in Theory and Practic...

2020
[15]

Yannick Hold-Geoffroy, Akshaya Athawale, and Jean-François Lalonde. 2019. Deep Sky Modeling For Single Image Outdoor Lighting Estimation. In CVPR

2019
[16]

Yannick Hold-Geoffroy, Kalyan Sunkavalli, Sunil Hadap, Emiliano Gambaretto, and Jean-François Lalonde. 2017. Deep outdoor illumination estimation. In Pro- ceedings of the IEEE conference on computer vision and pattern recognition . 7312– 7321

2017
[17]

Lukas Hosek and Alexander Wilkie. 2012. An Analytic Model for Full Spectral Sky-Dome Radiance. ACM Trans. Graph. 31, 4, Article 95 (jul 2012), 9 pages. https://doi.org/10.1145/2185520.2185591

2012
[18]

Lukáš Hošek and Alexander Wilkie. 2013. Adding a Solar-Radiance Function to the Hošek-Wilkie Skylight Model. IEEE Computer Graphics and Applications 33, 3 (2013), 44–52. https://doi.org/10.1109/MCG.2013.18

2013 doi
[19]

Lalonde, L.-P

J.-F. Lalonde, L.-P. Asselin, J. Becirovski, Y. Hold-Geoffroy, M. Garon, M.-A. Gard- ner, and J. Zhang. 2016. The Laval HDR sky database. http://sky.hdrdb.com

2016
[20]

Stark, Peter Shirley, and Simon Premože

Henrik Wann Jensen, Frédo Durand, Julie Dorsey, Michael M. Stark, Peter Shirley, and Simon Premože. 2001. A Physically-Based Night Sky Model. In Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques (SIGGRAPH ’01). Association for Computing ...

2001
[21]

Simon Kallweit, Thomas Müller, Brian McWilliams, Markus Gross, and Jan Novák
[22]

Kider, Daniel Knowlton, Jeremy Newlin, Yining Karl Li, and Donald P

Joseph T. Kider, Daniel Knowlton, Jeremy Newlin, Yining Karl Li, and Donald P. Greenberg. 2014. A Framework for the Experimental Comparison of Solar and Skydome Illumination. ACM Trans. Graph. 33, 6, Article 180 (nov 2014), 12 pages. https://doi.org/10.1145/2661229.2661259

2014
[23]

TL Koehler, RW Johnson, and JE Shields. 1991. Status of the whole sky imager database. In Proceedings of the Cloud Impacts on DOD Operations and Systems, 1991 Conference. 77–80

1991
[24]

Jean-François Lalonde and Iain Matthews. 2014. Lighting estimation in outdoor image collections. In 2014 2nd International Conference on 3D Vision , Vol. 1. IEEE, 131–138

2014
[25]

Martin Mirbauer, Tobias Rittig, Tomáš Iser, Jaroslav Krivánek, and Elena Šikudová
[26]

Parry Moon. 1940. Proposed standard solar-radiation curves for engineering use. Journal of the Franklin Institute 230, 5 (1940), 583–617. https://doi.org/10.1016/ S0016-0032(40)90364-7

1940
[27]

Tomoyuki Nishita, Yoshinori Dobashi, and Eihachiro Nakamae. 1996. Display of clouds taking into account multiple anisotropic scattering and sky light. In Proceedings of the 23rd annual conference on Computer graphics and interactive techniques. 379–386

1996
[28]

Tomoyuki Nishita, Takao Sirai, Katsumi Tadamura, and Eihachiro Nakamae. 1993. Display of the earth taking into account atmospheric scattering. In Proceedings of the 20th annual conference on Computer graphics and interactive techniques . 175–182

1993
[29]

Sean O’Neil. 2005. Accurate atmospheric scattering. Gpu Gems 2 (2005), 253–268

2005
[30]

Perez, R

R. Perez, R. Seals, and J. Michalsky. 1993. All-weather model for sky luminance distribution—Preliminary configuration and validation. Solar Energy 50, 3 (1993), 235–245. https://doi.org/10.1016/0038-092X(93)90017-I

1993 doi
[31]

A. J. Preetham, Peter Shirley, and Brian Smits. 1999. A Practical Analytic Model for Daylight. In Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques (SIGGRAPH ’99). ACM Press/Addison-Wesley Publishing Co., USA, 91–100. https://doi.org/10....

1999
[32]

Erik Reinhard, Wolfgang Heidrich, Paul Debevec, Sumanta Pattanaik, Greg Ward, and Karol Myszkowski. 2010. High dynamic range imaging: acquisition, display, and image-based lighting. Morgan Kaufmann

2010
[33]

Pinar Satilmis, Demetris Marnerides, Kurt Debattista, and Thomas Bashford- Rogers. 2022. Deep Synthesis of Cloud Lighting. IEEE Computer Graphics and Applications (2022)

2022
[34]

Chris Scullion. 2021. Forza Horizon 5’s art team explains how it got the skies to look so realistic. videogameschronicle.com [Online; posted 29-June-2021]

2021
[35]

Gowri Somanath and Daniel Kurz. 2020. HDR Environment Map Estimation for Real-Time Augmented Reality. CoRR abs/2011.10687 (2020). arXiv:2011.10687 https://arxiv.org/abs/2011.10687

2020 arXiv
[36]

Brandon Stafford. 2018. pysolar. https://doi.org/10.5281/zenodo.1461066

2018 doi
[37]

Jessi Stumpfel, Andrew Jones, Andreas Wenger, Chris Tchou, Tim Hawkins, and Paul Debevec. 2006. Direct HDR Capture of the Sun and Sky. In ACM SIGGRAPH 2006 Courses (Boston, Massachusetts) (SIGGRAPH ’06). Association for Computing Machinery, New York, NY, USA, 5–es. https://doi...

2006
[38]

The Advanced Rendering Toolkit. 2018. ART. https://cgg.mff.cuni.cz/ART/ archivers/art_2_1_1.html

2018
[39]

International Communications Union. 2015. Parameter values for the HDTV standards for production and international programme exchange. https://www. color.org/chardata/rgb/BT709.xalter

2015
[40]

Maquignaz, Hadi Moazen, Rishikesh Madan, Yannick Hold- Geoffroy, and Jean-François Lalonde

Lucas Valença, Ian J. Maquignaz, Hadi Moazen, Rishikesh Madan, Yannick Hold- Geoffroy, and Jean-François Lalonde. 2023. LM-GAN: A Photorealistic All-Weather Parametric Sky Model. https://doi.org/10.48550/ARXIV.2302.00087

2023 doi arXiv
[41]

Jianyi Wang, Kelvin C. K. Chan, and Chen Change Loy. 2022. Exploring CLIP for Assessing the Look and Feel of Images. arXiv:2207.12396 [cs.CV] https: //arxiv.org/abs/2207.12396

2022 arXiv
[42]

Alexander Wilkie and Lukas Hošek. 2013. Predicting Sky Dome Appearance on Earth-like Extrasolar Worlds. InProceedings of the 29th Spring Conference on Com- puter Graphics (Smolenice, Slovakia) (SCCG ’13). Association for Computing Ma- chinery, New York, NY, USA, 145–152. https...

2013
[43]

Alexander Wilkie, Petr Vevoda, Thomas Bashford-Rogers, Lukáš Hošek, Tomáš Iser, Monika Kolářová, Tobias Rittig, and Jaroslav Křivánek. 2021. A Fitted Radiance and Attenuation Model for Realistic Atmospheres. ACM Trans. Graph. 40, 4, Article 135 (July 2021), 14 pages. https://d...

2021
[44]

Piaopiao Yu, Jie Guo, Fan Huang, Cheng Zhou, Hongwei Che, Xiao Ling, and Yan- wen Guo. 2021. Hierarchical Disentangled Representation Learning for Outdoor Illumination Estimation and Editing. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 15313–15322

2021
[45]

Piaopiao Yu, Jie Guo, Longhai Wu, Cheng Zhou, Mengtian Li, Chenchen Wang, and Yanwen Guo. 2021. Dual attention autoencoder for all-weather outdoor lighting estimation. Sci. China Inf. Sci. 64 (2021)

2021
[46]

Jinsong Zhang, Kalyan Sunkavalli, Yannick Hold-Geoffroy, Sunil Hadap, Jonathan Eisenman, and Jean-François Lalonde. 2019. All-weather deep outdoor lighting estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 10158–10166

2019
[47]

Efros, Eli Shechtman, and Oliver Wang

Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang
[2008]

In Proceedings of the 2008 symposium on Interactive 3D graphics and games

Interactive multiple anisotropic scattering in clouds. In Proceedings of the 2008 symposium on Interactive 3D graphics and games . 173–182

2008
[2017]

ACM Trans

Deep Scattering: Rendering Atmospheric Clouds with Radiance-Predicting Neural Networks. ACM Trans. Graph. (Proc. of Siggraph Asia) 36, 6, Article 231 (Nov. 2017), 11 pages

2017
[2018]

https://doi.org/10.48550/ARXIV.1801.03924 Towards Physically-Based Sky-Modeling • 9 A FIGURES Fig

The Unreasonable Effectiveness of Deep Features as a Perceptual Metric. https://doi.org/10.48550/ARXIV.1801.03924 Towards Physically-Based Sky-Modeling • 9 A FIGURES Fig. 8. Images from modular training of latent2EDR 10 • Ian J. Maquignaz Fig. 9. HDRDB labels paired to fig. 8 ...

doi arXiv
[2022]

In Eurographics Symposium on Rendering

SkyGAN: Towards realistic cloud imagery for image based lighting. In Eurographics Symposium on Rendering. The Eurographics Association
[2049]

https://doi.org/10.5194/amt-11-2041-2018

2018 doi

Pith tools

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

[1] [1]

Khalid Alshaibani and Danny Li. 2021. Sky type classification for the ISO/CIE Stan- dard General Skies: a proposal for a new approach. International Journal of Low- Carbon Technologies 16, 3 (03 2021), 921–926. https://doi.org/10.1093/ijlct/ctab020 arXiv:https://academic.oup.com/ijlct/article-pdf/16/3/921/40490442/ctab020.pdf 8 • Ian J. Maquignaz

work page doi:10.1093/ijlct/ctab020 2021

[2] [2]

Antoine Bouthors, Fabrice Neyret, Nelson Max, Eric Bruneton, and Cyril Crassin

work page

[3] [3]

Eric Bruneton. 2017. A Qualitative and Quantitative Evaluation of 8 Clear Sky Models. IEEE Transactions on Visualization and Computer Graphics 23, 12 (2017), 2641–2655

work page 2017

[4] [4]

Eric Bruneton and Fabrice Neyret. 2008. Precomputed Atmospheric Scattering. Computer Graphics Forum 27, 4 (2008), 1079–1086

work page 2008

[5] [5]

Zhaoxi Chen, Guangcong Wang, and Ziwei Liu. 2022. Text2Light: Zero-Shot Text-Driven HDR Panorama Generation. ACM Transactions on Graphics (TOG) 41, 6, Article 195 (2022), 16 pages

work page 2022

[6] [6]

Paul Debevec. 1998. Rendering synthetic objects into real scenes: Bridging tra- ditional and image-based graphics with global illumination and high dynamic range photography. In Proc. SIGGRAPH 98. 189–198

work page 1998

[7] [7]

S. Dev, F. M. Savoy, Y. H. Lee, and S. Winkler. 2018. High-dynamic-range imaging for cloud segmentation. Atmospheric Measurement Techniques 11, 4 (2018), 2041–

work page 2018

[8] [8]

David Eigen, Christian Puhrsch, and Rob Fergus. 2014. Depth Map Prediction from a Single Image using a Multi-Scale Deep Network. arXiv:1406.2283 [cs.CV]

arXiv 2014