- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
This paper addresses 3D Gaussian scene reconstruction and novel view synthesis under challenging exposure and illumination conditions. It considers three settings: cross-view exposure variation (CEV), where exposure differs across images; spatial illumination variation (SIV), where different regions within an image receive uneven illumination; and high-contrast illumination (HCI), where bright regions coexist with deep shadows. The goal is to reconstruct a consistent, more evenly illuminated base appearance from these challenging observations.
To achieve this, the paper proposes a coupled 2D–3D optimization framework that jointly learns an image-space illumination decomposition and a Gaussian-level illumination field. The two branches constrain each other by aligning their illumination estimates and recombining the estimated base appearance with illumination to reconstruct the input images. Global camera-response correction and local exposure compensation further account for residual differences across training images. The resulting shared scene representation supports base-appearance rendering from novel viewpoints. Moreover, this paper presents a benchmark combining synthetic data with real-world data, covering diverse indoor and outdoor scenes under varying lighting conditions.
Critical strengths
- The coupled 2D–3D optimization is well motivated, combining complementary strengths of 2D and 3D representations while encouraging consistent illumination estimates.
- Global camera-response correction and ILEC address image-wide and local variations, with illumination-guided regularization constraining local adjustments. Jointly represent every capture-specific photometric residual
- Quantitative and qualitative results support the method’s effectiveness across datasets and challenging lighting conditions.
- Component ablations demonstrate the contributions of the main modules and regularization terms.
Critical weaknesses
-
The ablation studies of the 2D and 3D optimization modules are insufficient, as the key innovation lies in their mutual guidance and alignment. Additional experiments should be conducted to evaluate the effects of individual alignment and the absence of alignment.
-
The ground truth images in real-world scenarios only serve as references under specific conditions. Directly relying on quantitative metrics for evaluation may therefore be misleading. For example, reconstructed images may preserve the correct overall structure but differ slightly in global brightness (The overall brightness remains uniform but deviates from that of the GT images), leading to lower metric scores that do not accurately reflect the actual reconstruction quality.
-
The adaptive curve introduces a strong prior that may brighten inherently dark materials or darken inherently bright materials, potentially compromising the fidelity of material appearance.
Questions for authors
-
Could the authors provide additional ablation experiments to address the first issue raised in Critical Weaknesses Q1?
-
Could the authors propose a more appropriate evaluation method for real-world scenarios? Alternatively, if the current evaluation protocol is considered reasonable, could the authors provide further justification?
-
Could the authors further discuss the trade-off mentioned in Critical Weaknesses Q3 and analyze the scenarios in which it may lead to undesirable results?
-
Could the authors provide an analysis of the impact of the proposed module on training and rendering efficiency, including training time and rendering FPS?
5 comments
Rebuttal by Authors (continued)
W3 / Q3. The adaptive curve may brighten dark materials or darken bright ones.
Yes. The adaptive curve loss acts below and above , with soft targets of 0.45/0.55. However, the intrinsic ambiguity between reflectance and illumination may cause dark materials to be over-brightened and bright surfaces to be suppressed, as illustrated by the dark chair and white cabinet in DogRoom (Fig. 16, row 1); white preservation provides a soft constraint but cannot fully prevent these errors. This limitation is particularly evident under strong local illumination contrast, where the model may incorrectly attribute intrinsic material brightness to illumination variations. The table below further examines the sensitivity to and the weight of other priors.
Table: Parameter sensitivity analysis across CEV, SIV, and HCI scenes. Each entry reports PSNR/SSIM/LPIPS, averaged over the scenes in the corresponding category.
| Parameter | Setting | CEV | SIV | HCI |
|---|---|---|---|---|
| All parameters | EvenSplat (default) | 19.526/0.708/0.305 | 19.511/0.786/0.169 | 16.950/0.682/0.241 |
| Strongest baselines | (PPISP, PPISP, 3DGS+CHROMA) | 17.794/0.723/0.269 | 18.112/0.771/0.178 | 13.780/0.522/0.352 |
| (default = 0.5) | 0.35 | 20.016/0.720/0.301 | 19.868/0.800/0.165 | 16.716/0.671/0.246 |
| 0.65 | 18.413/0.745/0.269 | 19.285/0.779/0.170 | 17.234/0.691/0.239 | |
| (default = 1.0) | 0.5 | 19.284/0.709/0.298 | 19.269/0.787/0.167 | 17.268/0.685/0.245 |
| 1.5 | 18.957/0.691/0.303 | 19.700/0.787/0.170 | 16.801/0.675/0.244 | |
| (default = 0.2) | 0.1 | 19.515/0.714/0.303 | 19.508/0.786/0.169 | 16.902/0.679/0.244 |
| 0.3 | 19.893/0.714/0.297 | 19.519/0.787/0.169 | 17.107/0.686/0.240 |
Q4. Training time and rendering FPS.
The table below reports the computational cost on a single RTX 3090 GPU at 750×500 resolution. Due to the additional 2D decomposition network, illumination field, and image-specific correction modules, EvenSplat inevitably requires 2.21× the training time and 3.22× the peak training memory of 3DGS, trained for 10k iterations. At inference, the 2D network, CRN, and ILEC are discarded, while the illumination field is retained. EvenSplat achieves 74.36 FPS, compared with 289.60 FPS for 3DGS, maintaining interactive rendering performance.
| Method | train time (min) | peak mem (GB) | render FPS |
|---|---|---|---|
| 3DGS (gsplat) | 15.66 | 1.0428 | 289.60 |
| EvenSplat | 34.58 | 3.3546 | 74.36 |
Comment by Authors
As the author-reviewer discussion period comes to an end, we wanted to briefly check in to see whether our responses have resolved your concerns. If anything remains unclear, we would be happy to offer further clarification or additional analysis. We thank you again for your time and for your thorough feedback.
Rebuttal by Authors
We thank reviewer
3uVsfor the supportive review and the suggestions on evaluation.W1 / Q1. Ablate alignment individually and the absence of alignment.
We ran both variants: alignment only (λrecomb=0) and recombination only (λalign=0). Results are in the table below and in Appendix Table 12 of the revised paper. On HCI, alignment only and recombination only reach 15.70 and 15.83 dB, respectively, versus 16.95 dB for the full model. This shows that the two terms are complementary, and they impose different weightings. With R2D=I⊘L2D, the recombination residual is I⊙(Ladapt/L2D−1), weighting relative errors by image intensity, whereas alignment directly penalizes illumination discrepancies.
Removing both terms removes Lcouple entirely. Since Lcouple is the only path from the image-space branch to the Gaussians, this is the revised Table 5 row "No Image-Space Decomp.", which drops HCI PSNR to 12.01 dB. The coupling is bidirectional by design, and gradients from Lcouple update both the image-space network and the Gaussian illumination branch. We also test one-way coupling (stop-gradient on the 2D outputs), which lowers HCI PSNR to 15.18 dB.
Table: Coupling ablations on real scenes (PSNR/SSIM/LPIPS).
W2 / Q2. Real-world references' validity; a structurally correct but globally brighter/darker reconstruction is penalised. Justify the protocol.
We agree that absolute metrics might conflate global brightness differences with reconstruction errors, and therefore additionally report oracle per-image, per-channel affine-aligned metrics (yc=acxc+bc) on HCI dataset, as used for FourLogs in Fig. 11, to reduce global brightness and colour discrepancies (see the table below). After correction, EvenSplat remains ahead of the strongest baselines by 2.26 dB, also indicating that our HCI improvement cannot be explained solely by global brightness alignment. We retain uncorrected metrics as the primary evaluation in the paper because they directly measure reconstruction accuracy against the reference without GT-dependent post-processing, while the reference-free cross-lighting consistency test (Sec. 4.2) provides complementary evidence.