- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper solves the problem in distinguishing real land-cover changes from variations in remote sensing images. UCDNet uses cross-temporal co-relation to estimate shared features and subtracts these from the temporal representations to obtain the change features. A contrastive objective encourages discrimination between change and no-change, while a prediction head models class probabilities through a Dirichlet distribution and produces per-pixel uncertainty estimates. The authors motivate the feature decomposition through a structural causal model. Experiments cover LEVIR-CD, SYSU-CD, SECOND, and CDD-CD, with additional module ablations, cross-domain evaluation, complexity analysis, and feature-separability measurements.
Critical strengths
According to me the critical strengths are listed below :
-
Coherent Design Framework: Shared-feature suppression, prototype-based discrimination, and evidential prediction address some of the important aspects of change detection. The uncertainty maps shown in Figure 6 appear to highlight ambiguous regions and boundaries. Such outputs could be useful for identifying predictions that require further inspection, provided their relationship to prediction errors is quantitatively validated. Thus I feel it is interpretable.
-
The evaluation covers several complementary aspects of the method. Experiments span four datasets and eight comparison methods, alongside component ablations, cross-domain testing, complexity measurements, and feature-separability analysis. Reporting transfer limitations and computational cost is helpful because it gives a broader picture than in-domain accuracy alone.
Critical weaknesses
-
The causal interpretation is not established by the feature-subtraction argument. Equation 4 marginalizes over the proposed confounder, whereas the implemented method subtracts an estimated shared feature and predicts from the residual. These operations are not equivalent in general. The derivation should state which assumptions connect the learned shared feature to the proposed confounder and justify dropping the dependence on . For example, a conditional-independence or sufficiency condition would need to be stated and justified; it does not follow from subtraction alone. The first-order approximation and the cancellation of cross-terms also need a clearer validity argument, including the distribution underlying the expectations and any centering assumptions. The 0.80-point F1 drop when the disentanglement module is removed shows that the module is useful for prediction, but it does not distinguish causal deconfounding from ordinary nuisance suppression or feature regularization. A controlled intervention experiment and an ablation using simple feature subtraction would provide more direct evidence.
-
The experiments do not establish the reliability of the gains or isolate the claimed robustness mechanism. The improvements over the strongest baselines are 0.25 percentage points on LEVIR, 0.57 on SECOND, and 0.48 on CDD. Without repeated runs across seeds or uncertainty estimates, it is difficult to judge whether these differences are reproducible. This does not mean the gains are insignificant, but their statistical reliability remains unclear. The cross-domain results also do not show a consistent advantage: on SECOND, Table 5 reports 54.09 for UCDNet versus 57.66 for DMINet. Since temporal invariance and cross-domain invariance are different properties, this does not refute the causal model, but it leaves the claimed robustness benefit insufficiently supported. The authors should specify which variations the method is intended to remove and test them under controlled conditions. Reporting change-class IoU alongside mIoU would also make performance on the target class easier to interpret.
-
The uncertainty component needs quantitative validation beyond qualitative maps. Uncertainty concentrated near object boundaries is plausible, but it does not show that the predicted probabilities are calibrated or that the uncertainty reliably identifies errors. This matters because uncertainty estimation is presented as a central contribution. The evaluation should include calibration measures, such as reliability diagrams, ECE, Brier score, or negative log-likelihood, together with an error-ranking evaluation such as risk–coverage or sparsification curves. A comparison against predictive entropy from a standard classification head with the same backbone would help establish the benefit of the evidential formulation. The bounded activation also deserves an ablation and an explanation of how its per-stage evidence range interacts with the subsequent evidence fusion.
-
The contribution beyond UAFDcd needs to be made explicit. The published description of UAFDcd already includes a causal backdoor argument, correlation-based extraction of shared temporal features, differential feature separation, prototype contrastive learning, and Dirichlet-based uncertainty estimation [1]. Causal framing alone is therefore not enough to distinguish UCDNet. This overlap does not mean the methods are equivalent, but it calls for a precise comparison. The paper should identify the differences in the decomposition mechanism, prototype construction, contrastive objective, and uncertainty formulation. If maximum-entropy core-prototype selection is a main addition, comparisons against centroid, medoid, and other prototype-selection strategies would help establish its contribution. Controlled comparisons with matched backbones and training settings would be more informative than aggregate performance numbers alone.
Reference:
[1] B. Wang, S. Qiao, Z. Zhao, P. Qin, and J. Zeng. UAFDcd: Uncertainty-Aware Feature Decoupling Network for Remote Sensing Change Detection. IEEE Transactions on Geoscience and Remote Sensing, vol. 64, 2026. DOI: 10.1109/TGRS.2026.3701174.
Questions for authors
These are my questions for all the authors:
-
The decomposition in Equations 1–2 relies on assumptions about how the features separate. How are these assumptions connected to the learned ? Please also specify the expectations, centering assumptions, and approximation regime used in Equations 1–2.
-
Does in the causal graph denote the physical change label or the model's prediction? Do the other nodes represent physical scene factors or learned features? Please clarify how the arrows and interventions should be interpreted.
-
Does "invariant" mean shared across timestamps, insensitive to appearance perturbations, or transferable across datasets? Which of these properties does the analysis establish? Could controlled appearance changes, with the physical change mask held fixed, be used to test the claimed mechanism?
-
Could you report repeated-run results for Table 1, including the variability of the differences from the strongest baselines? Please clarify whether F1 is computed for the change class or macro-averaged, how scores are aggregated, and report change-class IoU alongside mIoU.
-
When is removed in Table 2, which features are supplied to , and what replaces the shared/private decomposition? Please also confirm the precision reported in the no-DCL row and explain the corresponding precision–recall trade-off.
-
Why is the maximum-entropy candidate preferable to a centroid or medoid as the core prototype? What benefit does this selection provide independently of the rest of the contrastive-learning design?
-
Why use for evidence generation? How does its bounded per-stage range affect the final fused evidence and uncertainty? Could you provide quantitative calibration and error-ranking results against a standard predictive-entropy baseline?
-
What are the specific methodological differences from UAFDcd? Please also clarify the number of evidence stages, the cluster counts, whether is fixed across datasets, and whether code and dataset-specific configurations will be released.
4 comments
Response to Reviewer 34sA (Part 2/4)
Q3:
Response:
Thank you for this critical suggestion.
- Conceptual Definition: In this work, "invariance" specifically refers to the insensitivity of temporal common-mode representations to irrelevant appearance perturbations (such as illumination gradients, seasonal phenology, and sensor gain variations), rather than cross-region or cross-resolution domain invariance.
- Controlled Appearance Perturbation Intervention Experiment: Following your suggestion, we constructed a controlled intervention benchmark on the LEVIR-CD test set: keeping the ground-truth change mask strictly fixed, we applied four representative non-causal appearance perturbations to the post-temporal imagery and compared UCDNet against a parameter-matched Non-Causal Baseline (replacing with stacked standard convolutional residual blocks of identical parameter budget). The results are summarized below:
| Intervention Condition | Non-Causal Baseline F1 / (%) | UCDNet F1 / (%) | (Base vs. Ours) | (Base vs. Ours) |
|---|---|---|---|---|
| Clean (Original) | 89.15 / 80.42 | 89.95 / 81.73 | 0.00% vs. 0.00% | 0.00% vs. 0.00% |
| Overexposure () | 88.40 / 79.21 | 89.26 / 80.61 | -0.75% vs. -0.69% | -1.21% vs. -1.12% |
| Shadow / Dark () | 87.18 / 77.28 | 88.50 / 79.37 | -1.97% vs. -1.45% | -3.14% vs. -2.36% |
| Seasonal Shift () | 84.47 / 73.12 | 85.76 / 75.07 | -4.68% vs. -4.19% | -7.30% vs. -6.66% |
| Sensor Noise () | 87.50 / 78.49 | 88.55 / 79.45 | -1.65% vs. -1.40% | -1.93% vs. -2.28% |
The results confirm that across all controlled appearance disturbances, UCDNet maintains superior absolute accuracy and exhibits substantially lower performance degradation under strong illumination, shadow, and seasonal shifts than the parameter-matched non-causal baseline, directly validating the causal disentanglement mechanism.
- Visualizing Controlled Nuisance Interventions: To visually illustrate the severity and physical realism of the synthetic perturbations, while keeping the pre-temporal image and ground-truth mask strictly fixed, we apply four non-causal appearance perturbations exclusively to (visualized in Fig. 5 of the revised manuscript):
- Overexposure: Simulates intense midday solar illumination and atmospheric scattering, markedly elevating overall brightness and washing out low-contrast surface textures.
- Shadow / Dark: Simulates cloud occlusion and dusk conditions, severely compressing radiometric dynamic range in dark regions.
- Seasonal Shift: Perturbs hue and saturation in the HSV space to replicate phenological vegetation color changes and spectral drift across seasons, altering surface colors without modifying physical building structures.
- Sensor Noise: Injects high-frequency additive Gaussian white noise to mimic sensor hardware degradation and low-SNR transmission interference.
Q4:
Response:
Thank you for your thorough requirements.
- Metric Definition & Aggregation Protocol: All F1, Precision, and Recall scores in the main text are strictly computed for the positive change class () rather than macro-averaged across background and change classes; mIoU is the mean of change-class IoU () and unchanged-class IoU ().
- Repeated Runs & Statistical Significance Analysis: We re-executed both UCDNet and the strongest baseline (UAFDcd [1]) across 5 identical random seeds on LEVIR-CD, reporting sample mean and standard deviation (), net gain (), two-tailed Welch's t-test -values, and Change-Class IoU ():
| Method | Precision (%) | Recall (%) | F1-Score (%) | OA (%) | mIoU (%) | Change IoU () (%) |
|---|---|---|---|---|---|---|
| UAFDcd [1] | 90.64 ± 1.06 | 87.83 ± 0.91 | 89.20 ± 0.16 | 98.92 ± 0.02 | 89.69 ± 0.14 | 80.51 ± 0.26 |
| UCDNet (Ours) | 91.43 ± 0.36 | 88.19 ± 0.22 | 89.78 ± 0.15 | 98.98 ± 0.02 | 90.19 ± 0.13 | 81.45 ± 0.25 |
| Net Gain () | +0.79% | +0.36% | +0.57% | +0.06% | +0.50% | +0.94% |
| -value |
(Note: * denotes , ** denotes , and *** denotes .)
- Key Conclusions: UCDNet consistently outperforms UAFDcd across all 6 metrics. All comprehensive segmentation indicators pass strict hypothesis testing: Change-Class increases by +0.94% (), F1-Score by +0.57% (), mIoU by +0.50% (), and OA reaches 98.98% (), confirming that the gains are statistically significant and reproducible rather than random fluctuations.
Response to Reviewer 34sA (Part 3/4)
Q5:
Response:
- Feature Replacement Strategy When Removing : When the Pearson covariance-based disentanglement module is removed (
w/o GFDin Table 2), to ensure strict alignment in parameter capacity and input structure for (consistent with the non-causal baseline in our controlled intervention experiments), we replace with parameter-matched stacked standard convolutional residual blocks. This baseline directly maps concatenated bitemporal features via standard residual convolutions to estimate without any cross-temporal channel covariance modeling. Subsequently, the exact same residual subtraction ( and ) is performed, and are fed into for supervision. - Typographical Erratum & Precision–Recall Trade-off: The Precision value
99.59%in thew/o LDCLrow of Table 2 in the initial manuscript was indeed a typographical transcription error (the actual recorded experimental value is91.88%), which we sincerely apologize for and have corrected in the revised manuscript. Without contrastive margin expansion (w/o LDCL), the model struggles to push apart ambiguous negative samples in feature space, causing decision boundaries to become overly conservative (predicting changes only in high-confidence core regions), which manifests as high Precision (91.88%) paired with a sharp drop in Recall (86.66%) and lower overall F1.
Q6:
Response:
Thank you for this thoughtful question. We address this from both geometric manifold motivation and standalone ablation experiments:
- Geometric Motivation: In RSCD, changed regions (e.g., farmland-to-factory, water-to-road) exhibit strongly multimodal heterogeneous distributions in feature space. Conventional centroids or medoids compute geometric centers that frequently fall into low-density transition zones between distinct change sub-modes ("false average samples"); pulling features toward such artificial centers over-shrinks decision boundaries and increases false negatives (lower Recall). Maximum-entropy core prototypes select real candidate samples within each sub-cluster that maximize mutual information and coverage uniformity relative to the overall distribution, avoiding the "mean trap" while suppressing boundary outliers.
- Standalone Ablation of Prototype Selection: Freezing all other modules and hyperparameters of UCDNet and varying only the prototype selection strategy on LEVIR-CD yields:
| Prototype Selection Strategy | Precision (%) | Recall (%) | F1-Score (%) | mIoU (%) | (%) | OA (%) |
|---|---|---|---|---|---|---|
| Cluster Centroid | 92.269 | 85.239 | 88.615 | 89.195 | 79.557 | 98.884 |
| Cluster Medoid | 93.069 | 84.647 | 88.658 | 89.237 | 79.627 | 98.897 |
| Max-Entropy Core Prototype (Ours) | 91.932 | 88.045 | 89.946 | 90.340 | 81.729 | 98.997 |
Simply switching the prototype selection rule improves Recall by +2.81% over Cluster Centroid and boosts F1-Score by +1.33% (Change by +2.17%), confirming its standalone contribution.
Response to Reviewer 34sA (Part 4/4)
Q7:
Response:
Thank you for this constructive question.
- Bounded Evidence Mechanism & Multi-Stage Fusion: In Evidential Deep Learning (EDL), unbounded activations (such as Softplus or unbounded exponential) are prone to exponential explosion when feature activations are strong, causing total Dirichlet strength to diverge and inducing severe overconfidence. By adopting bounded exponential mapping , single-stage evidence increments are strictly constrained within a stable bounded interval. This stage-wise bounded design ensures balanced weighting between shallow high-resolution edge textures and deep high-level semantics during cross-stage evidence accumulation, preventing large-norm deep responses from overwhelming shallow boundary uncertainty signals.
- Quantitative Calibration & Error-Detection Comparison: We added quantitative evaluations of Expected Calibration Error (ECE), Brier Score, Negative Log-Likelihood (NLL), and Error Detection AUROC on the LEVIR-CD test set, comparing our method against a matched-backbone standard classification head (Shannon predictive entropy) and an unbounded Softplus evidence variant:
| Method | ECE | Brier Score | NLL | Error Detection AUROC |
|---|---|---|---|---|
| Standard Head (Softmax Predictive Entropy) | 0.0132 | 0.0159 | 0.0781 | 0.9320 |
| UCDNet (Unbounded Softplus Evidence) | 0.0387 | 0.0158 | 0.0846 | 0.9763 |
| UCDNet (Ours, Multi-Stage Bounded Evidence) | 0.0131 | 0.0152 | 0.0692 | 0.9731 |
- Gains over Standard Predictive Entropy: Incorporating Dirichlet evidential modeling lowers Brier Score from 0.0159 to 0.0152 and NLL from 0.0781 to 0.0692, while boosting Error Detection AUROC from 0.9320 to 0.9731, demonstrating superior capability in identifying ambiguous boundary errors.
- Necessity of Bounded Activation: While unbounded Softplus achieves a slightly higher AUROC (0.9763), the lack of an upper bound causes severe overconfidence, inflating ECE nearly threefold (from 0.0131 to 0.0387) and worsening NLL to 0.0846. Our bounded design achieves optimal probability calibration (ECE 0.0131) alongside top-tier error-detection discrimination (0.9731).
Q8:
Response:
Thank you for your question. We clarify the core methodological differences, hyperparameter consistency, and open-source plan below:
- Methodological Comparison:
| Dimension | UAFDcd (IEEE TGRS 2026) | UCDNet (Ours) | Core Difference & Motivation |
|---|---|---|---|
| Disentanglement Mechanism | Gated separation in difference space | Cross-temporal channel Pearson covariance adaptive projection matrix | Explicitly models cross-channel second-order linear correlation filtering with stronger mathematical rigor |
| Prototypical Contrast | Conventional single-center static prototype alignment | Dynamic K-Means + Max-Entropy Core Prototype selection | Preserves complex multimodal manifold topology of real changes and prevents single-centroid representation collapse |
| Uncertainty Modeling | Single-stage discriminative head Dirichlet output | Cross-stage progressive multi-level evidence accumulation with bounded normalization | Synergistically fuses high-frequency spatial textures and high-level semantic evidence to calibrate boundary uncertainty |
- Hyperparameters & Open-Source Release: The number of stages (), cluster counts (), and five loss weights () are strictly fixed across all four datasets without dataset-specific tuning. Source code, pretrained weights, reproduction scripts, and full configuration files have been released via an anonymous GitHub repository.
Response to Reviewer 34sA (Part 1/4)
Q1:
Response:
Thank you for your insightful question. In Section 3.1 of the revised manuscript, we have explicitly clarified the mathematical link between the feature-separation assumptions and the learned feature Fp, along with its regime of validity:
Q2:
Response:
Thank you for pointing out the need to further clarify node semantics.