- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper analyzes the intrinsic relationship between RFO and key physical parameters in the photogrammetric process and tailors this analysis to the off-nadir remote sensing setting. Building on the photogrammetric analysis and insights from modern foundation models, the authors propose LoDEOT and equip it with a customized denoising training framework.
Critical strengths
The paper demonstrates a commendably cautious approach to acknowledging its limitations.
Critical weaknesses
- The lack of a coherent motivation–method–evidence chain makes it difficult for readers to understand why the proposed design is necessary to address a specific problem and whether the experimental results directly demonstrate that the design effectively resolves that problem.
- Due to the lack of the latest comparison methods, it is impossible to fully demonstrate the superiority of this method.
Questions for authors
Please refer to the weaknesses listed above.
4 comments
Response Q1
Q1: The lack of a coherent motivation–method–evidence chain makes it difficult for readers to understand why the proposed design is necessary to address a specific problem and whether the experimental results directly demonstrate that the design effectively resolves that problem.
Thank you for your feedback. Building upon the foundational concepts outlined in our Meta Response above (and in our Response to Reviewer fYgp), we have restructured our presentation and added targeted controlled experiments to establish a clear, unbroken motivation–method–evidence chain:
Step 1: From Core Observations to Architectural Design (Motivation Method)
Our architecture is directly motivated by three foundational observations established in the Introduction:
- Observation 1 (Implicit 3D Modeling in Transformers): Recent advances in 3D vision (e.g., AC3D) demonstrate that Transformer architectures naturally and implicitly parse and encode 3D imaging geometry during training, without requiring explicit channel-wise supervision or direct disentanglement into raw physical parameters.
- Observation 2 (Compact Geometric Parameterization in Off-Nadir Remote Sensing): Due to the high-altitude, near-parallel imaging characteristics of off-nadir satellite/aerial imagery, the 3D photogrammetric projection can be analytically simplified into a minimal 5-parameter representation .
- Observation 3 (Layer-Wise Denoising Dynamics): Deep Transformer decoders iteratively refine predictions through a layer-by-layer denoising process.
How These Observations Dictate Our Method:
- Why a Compact 5D Offset Token ()? Combining Observations 1 & 2, because both the physical projection and the network readout model the exact same geometric mapping, the token's degrees of freedom must satisfy to span the displacement space without geometric information loss. Thus, the 5D derivation acts as a principled geometric bottleneck—allowing us to decouple high-dimensional semantic/mask queries from a compact 5D recurrent offset state, where the Transformer implicitly encodes combinations of the underlying physical factors.
- Why Layer-Wise Denoising Supervision? Guided by Observation 3, injecting controlled additive perturbations into target-indexed offset tokens and applying layer-wise clean-target supervision trains the decoder to progressively denoise and refine the 5D geometric state across layers rather than relying on single-step final-layer regression.
Step 2: Explaining Empirical Dimensional Behavior Across Three Regimes (Method Evidence)
To connect our theoretical 5D bound to the empirical results in Table 3 (where neighboring dimensions perform comparably on constrained subsets), we analyze the token width across three distinct regimes:
- 1. High-Dimensional Regime (): Prior baselines uniformly default to over-parameterized embeddings (). Our work breaks this convention by demonstrating that geometric offset regression can be solved within a compact low-dimensional subspace.
- 2. Over-Parameterized Low-Dimensional Regime (): In Table 3, accuracy across and shows no statistically significant variance because once , the token already satisfies the full-rank DoF bound (). Increasing beyond 5 merely introduces redundant latent capacity without new geometric information.
- 3. Sub-Rank Regime () — Conditional Dimensionality Reduction: Real-world benchmarks cover a finite set of satellite sensors, and off-nadir viewing angles are bounded within narrow operational ranges (e.g., , far from horizon grazing angles). Substituting real satellite parameters into our equations confirms conditional dimensionality reduction, collapsing the effective degrees of freedom from 5D onto an approximate 2D–2.5D manifold. Hence, 2D–2.5D reflects the dataset-specific numerical approximation, whereas 5D guarantees full-rank geometric completeness across unconstrained viewing geometries.
(Quantitative tables verifying the physical sensitivity, SVD spectral alignment, 5D vs. Direct 2D head, and layer-wise denoising ablation follow in Part 2/2 below.)
Response Q1
To directly substantiate every link in the chain above, we added four controlled quantitative experiments to the revised Appendix:
1. Real-World Satellite Physical Sensitivity Verification Substituting real orbital/camera parameters from five satellites (Jilin-1 GF03D, QuickBird, WorldView-2, WorldView-3, Sentinel-2) with empirical building heights (–), focal anisotropy (), normalized skew (), and viewing angles () into yields sensitivity indices (px):
| Satellite | (Height) | (Geometric) | ||||
|---|---|---|---|---|---|---|
| Jilin-1 GF03D | 0.0224 | 0.0250 | 3.3298 | 2.8997 | 2.0072 | 2.2434 |
| QuickBird | 0.0259 | 0.0288 | 3.8421 | 3.3458 | 2.3160 | 2.5886 |
| WorldView-2 | 0.0366 | 0.0407 | 5.4288 | 4.7276 | 3.2725 | 3.6577 |
| WorldView-3 | 0.0542 | 0.0604 | 8.0558 | 7.0153 | 4.8560 | 5.4276 |
| Sentinel-2 | 0.0006 | 0.0007 | 0.0854 | 0.0745 | 0.0493 | 0.0532 |
- Physical Finding: Height-related and geometric magnitude dominate intrinsic shape terms by two orders of magnitude, proving conditional collapse onto an approximate 2D–2.5D manifold under bounded viewing angles.
2. SVD Spectral Distribution & Principal Component Truncation Performing SVD on learned -dimensional offset features before the readout layer and truncating minor components at inference time shows:
| Dim | Top-1 Energy (%) | Top-2 Energy (%) | Full mEPE (px) | Top-1 Only mEPE (px) | Top-2 Only mEPE (px) | |
|---|---|---|---|---|---|---|
| 2 | 80.90 | 100.00 | 2 | 2.8239 | 3.2868 | 2.8239 |
| 3 | 87.70 | 99.998 | 2 | 2.7675 | 3.5467 | 2.7665 |
| 4 | 94.94 | 99.991 | 2 | 2.7563 | 3.3743 | 2.7574 |
| 5 | 80.93 | 99.992 | 2 | 2.6723 | 3.4273 | 2.6745 |
| 16 | 80.13 | 99.973 | 2 | 2.6946 | 3.3015 | 2.6958 |
| 256 | 79.51 | 99.960 | 2 | 2.6417 | 3.2986 | 2.6819 |
- Spectral Finding: For all , the top-2 components capture 99.960%–99.998% variance (). For our 5D model, dropping components 3–5 changes mEPE by only (), while dropping the 2nd component degrades mEPE to , matching the 2 dominant physical factors ().
3. Direct Attribution to the 5D Offset Representation (vs. Same-Framework Direct 2D Head) To isolate the exact gain of our 5D geometric bottleneck from backbone or training recipes, we replaced our offset branch with a standard Direct 2D Regression Head within the exact same framework and training schedule:
| Offset Head Design | Footprint AP | Footprint | Offset mEPE (px) |
|---|---|---|---|
| Direct 2D Regression Head | 19.79 | 42.08 | 3.21 |
| LoDEOT (5D Offset Token, Ours) | 22.72 | 47.22 | 2.81 |
| Net Gain () | +2.93 | +5.14 | -0.40 (-12.46%) |
- Finding: With all other variables held strictly constant, our 5D offset representation alone improves Footprint by and reduces offset error by .
4. Independent Ablation of Layer-Wise Denoising Supervision To verify that our layer-wise denoising framework (Observation 3) actively resolves iterative error accumulation, we independently varied supervision depth and perturbation type under the exact same backbone:
| Supervision Depth | Perturbation Type | Footprint AP | Footprint | Offset mEPE (px) |
|---|---|---|---|---|
| Final-layer | None | 20.89 | 43.84 | 3.30 |
| Final-layer | Additive | 20.42 | 42.88 | 3.32 |
| Layer-wise | None | 19.42 | 42.01 | 3.30 |
| Layer-wise | Multiplicative | 21.04 | 44.16 | 3.02 |
| Layer-wise | Additive (Ours) | 22.72 | 47.22 | 2.81 |
- Finding: Adding noise only at the final layer degrades performance, whereas pairing layer-wise supervision with additive noise reduces mEPE by () over the un-noised baseline.
Together, these four controlled experiments establish a complete, unbroken chain from physical observation to architectural design and empirical improvement.
Response Q2
Q2: Due to the lack of the latest comparison methods, it is impossible to fully demonstrate the superiority of this method.
To the best of our knowledge, Table 2 in the main text already evaluates all publicly available, reproducible open-source baselines in off-nadir building footprint extraction (including externally guided methods such as OBM and PolyFootNet, and end-to-end methods such as LOFT, Cas.LOFT, and MLS-BRN).
We also thoroughly examined the most recent literature in this area—specifically DG-BRF (ISPRS J. Photogramm. Remote Sens., 2026) and DragRoof (arXiv:2607.25210, 2026)—and checked their official repositories prior to submission and during the rebuttal period:
- DG-BRF (
https://github.com/zhenghaohu/DG-BRF): Although the paper was published earlier this year, its official repository has not yet released any source code or pretrained weights (community requests in its GitHub Issues remain open). - DragRoof (
https://github.com/likaiucas/DragRoof): Currently a preprint whose official repository similarly lacks public code or model checkpoints.
Because neither method provides official code or pretrained weights, re-implementing them from scratch without verified architectural details or hyperparameters would risk an unfair comparison. Should the authors release their official implementations—or if there are any other open-source, reproducible baselines we should include, we would be very glad to add them to our benchmark comparisons.
In other words, if there are any other open-source, reproducible baselines we may have overlooked, we would be grateful if you could point us to them.
Meta Response
Meta Response:
We sincerely thank you for dedicating your valuable time to reviewing our manuscript. We deeply regret that our current presentation did not clearly convey the connection between our underlying motivation and empirical validation. We take full responsibility for this clarity gap in our writing—as reflected by the fact that while Reviewer hVff grasped the geometric motivation and focused mainly on experimental verification, our exposition left these links insufficiently transparent in your reading.
We also appreciate your note regarding our discussion of the paper's limitations ("The paper demonstrates a commendably cautious approach to acknowledging its limitations"). To avoid any potential misunderstanding, we would like to clarify that this work is entirely human-conceived and executed—not an auto-research or AI-generated study (AI tools were used strictly for minor language polishing)—and our cautious tone stems solely from a commitment to scientific rigor rather than defensiveness.
Because our approach bridges two distinct subfields, fully appreciating the motivation requires combining domain knowledge in remote sensing segmentation with recent foundational findings in general domain, e.g., 3D vision and Video World Models (specifically, how Transformer architectures implicitly parse and encode 3D camera geometry without explicit physical supervision). To make this cross-domain foundation clear and accessible, we have summarized these core premises in our Meta Response to Reviewer fYgp and added a dedicated Related Work section in the revised manuscript. We kindly invite you to refer to that overview alongside our point-by-point responses below.