- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
The authors propose GOLD, a diffusion-based framework for domain-generalized object detection that formulates cross-domain localization as semantics-preserving proposal denoising. Built on DiffusionDet, GOLD combines SCAdapter for spatially calibrated DINOv3 features, SHEP for region-dependent visual perturbations, and PAD for prototype-guided proposal refinement. Experiments on Diverse Weather, Cityscapes→BDD100K, and Cityscapes→FoggyCityscapes demonstrate consistent improvements over existing DGOD methods.
Critical strengths
- The paper proposes GOLD, a diffusion-based DGOD framework built on DiffusionDet. It combines SCAdapter for spatially adapted DINOv3 features, SHEP for region-specific perturbations, and PAD for multi-scale prototype guidance during proposal refinement.
- The method is evaluated on Diverse Weather, Cityscapes→BDD100K, and Cityscapes→Foggy Cityscapes. Ablation studies examine the contributions of SCAdapter, SHEP, and PAD, as well as perturbation strengths and prototype design choices. The paper also reports training overhead and inference cost.
- The method and optimization objectives are clearly described, with pseudocode and hyperparameter tables provided. Stage-wise box refinement, Grad-CAM, and t-SNE visualizations further support the analysis. The work focuses on maintaining reliable spatial evidence during iterative localization under domain shifts.
Critical weaknesses
- SCAdapter is motivated by the claim that frozen DINOv3 features exhibit a mismatch between semantic representation and object localization. However, the paper does not provide direct experimental evidence to verify this claim or isolate the localization-specific contribution of SCAdapter.
- SHEP applies several perturbations to the input images, but real-world visual perturbations are diverse and difficult to exhaustively cover during training, including snow, spatter, frost, shot noise, impulse noise, speckle noise, motion blur, JPEG compression, pixelation, and geometric deformation. The paper does not evaluate GOLD on such unseen perturbations, leaving its robustness under out-of-distribution degradations unclear.
- The scalability of GOLD beyond the frozen DINOv3 backbone has not been demonstrated. Without experiments using alternative visual encoders or architectures, it remains unclear whether the reported improvements are generally applicable or primarily dependent on DINOv3’s pretrained representations.
Questions for authors
- SCAdapter is motivated by the claim that frozen DINOv3 features exhibit a mismatch between semantic representation and object localization. Is there experimental evidence to support this claim?
- SHEP applies several perturbations to the input images, but real-world visual perturbations are diverse and difficult to exhaustively cover during training, including unseen cases such as snow, spatter, frost, shot noise, impulse noise, speckle noise, motion blur, JPEG compression, pixelation, and geometric deformation. How well does the proposed method perform when directly applied to test images with these unseen perturbations? This would provide a more intuitive evaluation of GOLD’s actual contribution.
- The scalability of the method to other architectures has not yet been demonstrated. If DINOv3 is replaced with other visual encoders, how much performance improvement can the method still achieve?
5 comments
Response to Reviewer KGcU Weakness 2 & Question 2
KGcU - Weakness 2 & Question 2
The public imagecorruptions implementation provides standardized noise, blur, weather, and digital corruptions, including snow, spatter, frost, shot noise, impulse noise, speckle noise, motion blur, JPEG compression, and pixelation. We are conducting an additional evaluation on Cityscapes-Corruption using models trained on Cityscapes:
Table KGcU 2. Generalization results on Cityscapes-Corruption. All models are trained on Cityscapes. For each type of corruption, the average mAP is reported over five severity levels, and mPC denotes the mean value over all types.
| Method | Venues | Noise | Blur | Weather | Digital | mPC | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gauss | Shot | Impulse | Defocus | Glass | Motion | Zoom | Snow | Frost | Fog | Bright | Contrast | Elastic | Pixel | JPEG' | |||
| StylizedAug | ICLR'19 | 4.8 | 6.8 | 4.3 | 19.5 | 18.7 | 18.5 | 2.7 | 3.5 | 17.0 | 30.5 | 31.9 | 22.7 | 33.9 | 22.6 | 20.8 | 17.2 |
| AutoAug | ECCV'20 | 0.9 | 1.6 | 0.9 | 16.8 | 14.4 | 18.9 | 2.0 | 1.9 | 16.0 | 32.9 | 35.2 | 26.3 | 39.4 | 17.9 | 11.6 | 15.8 |
| AugMix | ICLR'20 | 5.0 | 6.8 | 5.1 | 18.3 | 18.1 | 19.3 | 6.2 | 5.0 | 20.5 | 31.2 | 33.7 | 25.6 | 37.4 | 20.3 | 19.6 | 18.1 |
| SupCon | NeurIPS'20 | 7.0 | 9.5 | 7.4 | 22.6 | 20.2 | 22.3 | 4.3 | 5.3 | 23.0 | 37.3 | 38.9 | 31.6 | 40.1 | 24.0 | 20.1 | 20.9 |
| FSCE | CVPR'21 | 7.4 | 10.2 | 8.2 | 23.3 | 20.3 | 21.5 | 4.8 | 5.6 | 23.6 | 37.1 | 38.0 | 31.9 | 40.0 | 23.2 | 20.4 | 21.0 |
| OA-Mix | AAAI'24 | 7.2 | 9.6 | 7.7 | 22.8 | 18.8 | 21.9 | 5.4 | 5.2 | 23.6 | 37.3 | 38.7 | 31.9 | 40.2 | 22.2 | 20.2 | 20.8 |
| OA-DG | AAAI'24 | 8.2 | 10.6 | 8.4 | 24.6 | 20.5 | 22.3 | 4.8 | 6.1 | 25.0 | 38.4 | 39.7 | 32.8 | 40.2 | 23.8 | 22.0 | 21.8 |
| PhysAug | AAAI'25 | 14.3 | 17.0 | 11.9 | 25.6 | 19.1 | 25.5 | 3.9 | 8.6 | 21.3 | 35.3 | 39.5 | 27.5 | 39.1 | 28.9 | 19.9 | 22.6 |
| OPCD | MM'25 | 18.0 | 20.1 | 13.7 | 25.1 | 20.5 | 24.9 | 4.0 | 13.2 | 22.8 | 39.0 | 39.1 | 31.9 | 37.3 | 26.2 | 20.1 | 23.7 |
| UICL | AAAI'26 | 28.5 | 30.9 | 29.2 | 31.9 | 21.6 | 26.0 | 14.6 | 13.6 | 37.2 | 44.5 | 40.5 | 38.2 | 41.6 | 38.5 | 25.8 | 31.5 |
| Ours | - | 30.8 | 32.6 | 32.0 | 33.8 | 24.0 | 27.6 | 16.8 | 16.2 | 39.0 | 47.0 | 42.0 | 40.9 | 43.7 | 40.4 | 28.1 | 33.0 |
As shown in Table KGcU 2, GOLD achieves the highest mPC of 33.0%, outperforming the strongest baseline (UICL) by 1.5% mAP. Consistent improvements are observed across all 15 corruption types, spanning noise, blur, weather, and digital degradations. These results demonstrate that GOLD maintains strong robustness against diverse out-of-distribution perturbations, including those not explicitly considered during training. We also incorporate these results into the experiment section of the revision as Table 12 to further demonstrate GOLD's robustness against unseen visual perturbations (highlighted in green).
Response to Reviewer KGcU Weakness 3 & Question 3
KGcU - Weakness 3 & Question 3
We evaluate GOLD with ResNet-50, ResNet-101, Swin-B, and DINOv3 in DiffusionDet, matching the pretrained checkpoint and training settings within each baseline–GOLD pair. Table KGcU 3 reports both per-domain performance and the gain in four-domain average mAP.
Table KGcU 3. Cross-encoder evaluation on Diverse Weather.
| Visual encoder | Method | DF | DR | NR | NS | Avg. | Gain |
|---|---|---|---|---|---|---|---|
| ResNet-50 | Baseline | 36.5 | 29.0 | 15.7 | 36.8 | 29.5 | — |
| ResNet-50 | GOLD | 43.0 | 43.1 | 26.9 | 46.8 | 40.0 | +10.5 |
| ResNet-101 | Baseline | 38.1 | 30.8 | 17.3 | 38.2 | 31.1 | — |
| ResNet-101 | GOLD | 43.7 | 43.9 | 28.4 | 47.7 | 40.9 | +9.8 |
| Swin-B | Baseline | 42.3 | 49.2 | 28.7 | 51.2 | 42.9 | — |
| Swin-B | GOLD | 51.5 | 57.2 | 44.4 | 55.7 | 52.2 | +9.3 |
| DINOv3 | Baseline | 44.6 | 52.3 | 35.6 | 51.6 | 46.0 | — |
| DINOv3 | GOLD | 55.6 | 62.3 | 48.8 | 60.5 | 56.8 | +10.8 |
Replacing DINOv3 with ResNet-50, ResNet-101, or Swin-B still yields 10.5%, 9.8%, and 9.3% mAP gains, respectively, with improvements on every target domain. DINOv3 gains 10.8 points and achieves the highest final average mAP (56.8%). These results show that GOLD's benefits extend across the evaluated CNN and Transformer encoders, while encoder choice continues to affect the final performance. We have added these results and a detailed analysis to Appendix F, highlighted in green.
Response to Reviewer KGcU Weakness 1 & Question 1
KGcU - Weakness 1 & Question 1
A. Clarification of SCAdapter's Motivation.
Thank you for encouraging us to clarify the motivation and supporting evidence. We will revise the wording in lines 169–172, with the changes highlighted in green, to emphasize task-specific spatial adaptation rather than a general limitation of DINOv3 representations:
We have also revised the corresponding descriptions in the second contribution and the Conclusion to maintain consistent wording throughout the manuscript. These changes are highlighted in green.
B. Controlled Ablations of SCAdapter.
Importantly, Table 9 (rows 1 and 5) already provides a controlled comparison: with the same DINOv3 backbone and SHEP/PAD enabled, replacing linear adaptation with SCAdapter increases the average mAP from 53.5% to 56.8% (+3.3% mAP). We further isolate SCAdapter without SHEP/PAD in Table KGcU 1: rows 1 and 2 use the same frozen DINOv3 backbone, and SCAdapter improves the average from 46.0% to 48.8% (+2.8% mAP). All of these results can prove that SCAdapter helps improve the positioning capability of the model.
Table KGcU 1. Comparison of linear adaptation and SCAdapter on Diverse Weather, without SHEP or PAD.