M4-SAR: A Multi-Resolution, Multi-Polarization, Multi-Scene, Multi-Source Dataset and Benchmark for optical-SAR Object Detection

Chao Wang; Jian Yang; Lei Luo; Wei Lu; Xiang Li

REVIEW 4 major objections 6 minor 1 cited by

M4-SAR: A Multi-Resolution, Multi-Polarization, Multi-Scene, Multi-Source Dataset and Benchmark for optical-SAR Object Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A new benchmark with 981,862 labeled optical-SAR instances shows fusion beats either sensor alone.

desk verdict Genuinely useful new optical-SAR fusion benchmark with a directionally solid result, but the headline 5.7% fusion-gain figure is untraceable and the label-transfer step needs validation before the numbers can be trusted. read the letter →

arxiv 2505.10931 v4 pith:JLIXESUN submitted 2025-05-16 cs.CV

Chao Wang , Wei Lu , Xiang Li , Jian Yang , Lei Luo This is my paper · ORCID

classification cs.CV

keywords optical-SARfusionobjectdetectionrotatedboundingboxesremotesensingdatasetmulti-sourcesyntheticapertureradarSentinel-1Sentinel-2

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

This paper argues that optical-SAR fusion detection has been blocked by the absence of a large, standardized dataset, and it supplies one. M4-SAR contains 112,174 aligned optical-SAR image pairs and 981,862 rotated-box instances across six coastal-infrastructure categories, drawn from Sentinel-1 and Sentinel-2 imagery at two resolutions. Labels are produced on cloud-free optical images and transferred to the paired radar images and to cloudy optical replacements, keeping annotation costs down. On this benchmark every fusion detector tested beats its single-modality counterpart, and the proposed E2E-OSDet reaches 61.4% mAP; the paper reports a 5.7-point gain from fusing the two modalities. A released toolkit gives future methods a common protocol for comparison.

What carries the argument

The machinery is two coupled pieces. First, the dataset pipeline: a semi-supervised optical-assisted annotation loop in which a detector trained on 5% manually labeled cloud-free optical images produces pseudo-labels for the next 5%, humans correct them, and the cycle repeats; because optical and radar images are aligned only by geographic coordinates, the optical boxes are then transferred to the SAR pairs and to cloudy optical replacements. Second, E2E-OSDet, whose three modules each attack the optical-SAR domain gap: the Filter Augment Module projects both modalities through classical filters (HOG, Canny, Haar, gradient, wavelet scattering) to raise their structural similarity; the Cross-modal Mamba Interaction Module interleaves the two feature sequences so that a Mamba scan—a linear-time state-space sequence model—sees corresponding optical and SAR positions as adjacent; and the Area-attention Fusion Module weights salient local regions in the merged features before the oriented-box head.

What would settle it

Manually re-annotate a random sample of SAR test images—say 1,000 instances—and measure the average IoU between the fresh boxes and the labels transferred from optical images. If the mean IoU is well below 0.5, the ground truth is systematically displaced and the reported mAP values, including the 5.7-point fusion gain, rest on misaligned labels.

Watch

Extended reading notes

Core claim

The paper's central claim is that a large, instance-level aligned optical-SAR dataset is feasible to build by annotating the easier modality and transferring labels across sensors, and that such a dataset reveals a consistent, exploitable fusion advantage. M4-SAR covers bridges, harbors, oil tanks, playgrounds, airports, and wind turbines at 10 m and 60 m resolution, in VH and VV polarizations, with optical and radar acquisitions up to ten days apart. Across the benchmark, every fusion method beats its single-modality counterpart at the same backbone, with the largest gains in cloudy, snowy, low-light, and low-resolution scenes. E2E-OSDet, which narrows the cross-modal domain gap before fusing features, reports the top mAP of 61.4% on M4-SAR and the best results on the OGSOD-1.0 and OGSOD-2.0 benchmarks, supporting the claim that explicit domain alignment is what makes optical-SAR fusion work.

Load-bearing premise

The load-bearing premise is that labels drawn on clear optical images remain accurate enough when shifted, using only geographic-coordinate alignment, onto radar images taken up to ten days apart—even though the two modalities register poorly at the pixel level.

Editorial extensions

If this is right

Fusion beats single-source detection at fixed architecture: with YOLOv11 as the shared backbone, every fusion method in the benchmark exceeds both the optical-only (54.3 mAP) and SAR-only (44.7 mAP) results, and the gap is largest in degraded scenes.
The dataset's rotated boxes, dense instances (8.75 per image), category imbalance (bridges are 43.6% of instances), and extreme aspect-ratio spread make it a stress test for oriented detection under realistic coastal conditions.
Each of the three E2E-OSDet modules adds measurable gain in the ablation (from 58.0 to 61.4 mAP), and the full model transfers to other optical-SAR benchmarks, so aligning the feature domain before fusion is a generalizable design rather than a dataset-specific trick.
Releasing the dataset and toolkit with a shared backbone and evaluation protocol makes head-to-head comparisons of future optical-SAR detectors possible and reproducible.

Reading between the lines

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

Testable extension the authors do not run: split the optical test images into cloud-free and cloudy subsets and measure the fusion gain on each; the paper predicts the gain concentrates in degraded scenes, and that prediction is directly checkable with the released pairs.
The interleaved-sequence arrangement in CMIM is sensor-agnostic: the same optical-radar-optical-radar ordering could be applied to any poorly aligned multimodal pair, such as visible/infrared or multi-date radar, where forcing corresponding patches to be adjacent may help any sequence model stay aligned.
Because the categories are static coastal infrastructure, the alignment premise is untested on moving targets such as ships; a dynamic-category extension would be the sharpest stress test, since the up-to-ten-day acquisition gap would likely break label transfer there.
The headline 5.7-point fusion gain is not read directly off the main table: the gap to the best optical-only method is 5.1 points and to the same-backbone YOLOv11 optical result is 7.1 points, so pinning the single-source reference to one named method would make the headline number reproducible.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Two things to know up front. The dataset is real and fills a genuine gap: no prior optical-SAR benchmark combines this scale (112,174 paired images, ~982k instances), rotated boxes, six categories, dual resolution, and dual polarization, with test sets in both modalities. OGSOD's test sets are SAR-only and the large SAR datasets are single-source. Second, the headline number—5.7% fusion gain—is not traceable in Table 2. E2E-OSDet's 61.4 mAP beats the best optical-only YOLO by 5.1 points and the best SAR-only by 12.9, so the 5.7 as written does not align with any row of the table, and the comparison is confounded: MMRotate single-source baselines get 72 epochs with an ImageNet-pretrained ResNet-50, while fusion models get 300 epochs from scratch with a YOLOv11-S backbone. Fusion winning despite the handicap strengthens the qualitative conclusion, but the magnitude is unestablished.

What the paper does well: every one of the nine fusion methods beats every single-source method in Table 2, so the claim that fusion helps is directionally solid and not dependent on the authors' own detector. The ablations in Table 3 show each module contributing, Fig. 5 shows the framework is not fragile to hyperparameter choices, and the deliberate restriction to static targets is the right response to temporal misalignment. The limitations section is honest, including the weak-alignment numbers.

The load-bearing soft spot is the annotation pipeline. Labels are drawn on cloud-free optical images and transferred to SAR and to cloudy optical replacements via geographic-coordinate alignment that the paper itself rates as weak—SSIM 0.1256, up to 10-day acquisition gaps. No annotation-agreement or localization-error study is reported. The iterative pseudo-labeling also begins from a detector trained and evaluated on the same 5% subset, so detector errors can propagate into most of the 982k instances. The static-target constraint and the manual refinement step in each iteration blunt the worst case, but the transferred SAR labels are simply unvalidated, and for small classes like wind turbines (average ~96 px) a systematic offset would move rankings. This is the most serious concern and it is fixable: a sampled human re-annotation of SAR boxes, with agreement statistics, would largely settle it.

Smaller issues: the training conditions for the single-source YOLOs are not stated as explicitly as for the fusion models; hyperparameters appear selected on test numbers; and there are no error bars anywhere.

Bottom line: this paper deserves serious peer review. It is a substantial resource for the optical-SAR detection community, the flaws are addressable rather than structural, and the core claim holds directionally. I would send it out and require the annotation-quality study, an equalized baseline comparison, and an honest statement of where 5.7% comes from before acceptance.

Referee Report

4 major / 6 minor

Summary. The manuscript introduces M4-SAR, a paired optical-SAR rotated object detection dataset built from Sentinel-1 and Sentinel-2 data, containing 112,174 image pairs and 981,862 oriented instances over six static coastal infrastructure categories, with multi-resolution (10m/60m), multi-polarization (VV/VH), and multi-scene coverage. It also presents MSRODet, a unified benchmarking toolkit implementing seven existing fusion detectors, and proposes E2E-OSDet, an end-to-end fusion detector combining a Filter Augment Module (FAM), a Cross-modal Mamba Interaction Module (CMIM), and an Area-attention Fusion Module (AFM). On the M4-SAR test set, E2E-OSDet reports 61.4% mAP, and the paper claims a 5.7% mAP improvement from optical-SAR fusion over single-source inputs.

Significance. If the dataset labels are trustworthy, M4-SAR is a valuable community resource: it is substantially larger than existing optical-SAR paired benchmarks (OGSOD-1.0/2.0), adds rotated boxes and six categories, and ships with a standardized, reproducible evaluation toolkit and a strong baseline detector. The authors also provide public code and data. However, the benchmark's scientific value hinges on the validity of transferring optical annotations to SAR and cloudy images, which is asserted rather than demonstrated; the reported cross-modal alignment is weak (SSIM 0.1256, MSE 0.0662, MI 0.1729), so label noise could affect every reported mAP. The comparison protocols also mix training budgets and backbones across single-source baselines, making the headline fusion gain hard to interpret.

major comments (4)

[§2.4, Fig. 3(d), §5.1] The dataset's validity rests on the assertion in §2.4 that annotations derived from cloud-free optical images can be 'reliably transferred' to paired SAR images and to cloudy optical replacements, but no quantitative validation of this transfer is provided. The paper itself reports weak alignment (SSIM 0.1256, MSE 0.0662, MI 0.1729 in Fig. 3(d)) and acquisition time gaps of up to 10 days, and §5.1 concedes that alignment is geographic-coordinate-based rather than pixel-level. Because every mAP in Table 2, including E2E-OSDet's 61.4%, is computed against labels inherited from optical images, a systematic offset in SAR label placement—plausible at 10–60 m resolution, and especially damaging for the smallest class (wind turbines, average area about 96 px in Fig. 3(c))—would change all reported numbers, including the claimed fusion benefit. Please add an annotation-agreement study, a localization-error measurement on SAR, or a SAR-native manual verification subset to quantify label transfer accuracy.
[Abstract, Table 2] The abstract's headline claim that fusing optical and SAR data improves mAP by 5.7% over single-source inputs is not traceable to Table 2. The best optical-only result is YOLO-Master at 56.3% and the best SAR-only result is YOLO-Master at 48.5%, while the best fusion result (E2E-OSDet) is 61.4%, giving gaps of 5.1 and 12.9 points respectively; no pairwise combination yields 5.7. Please specify the exact baseline configuration and protocol underlying the 5.7% figure, or revise the abstract to report the numbers that Table 2 actually supports.
[§4.1, §10.1, Table 2] The comparison between multi-source fusion methods and the MMRotate single-source baselines is confounded by differing training protocols. All fusion methods are trained from scratch for 300 epochs with a YOLOv11-S backbone and YOLOv8 OBB head (§4.1, §10.1), whereas the MMRotate baselines (R-FCOS, R-ATSS, O-RepPoint, RTMDet, PSC, LSKNet) are trained for 72 epochs with an ImageNet-pretrained ResNet-50 under a different optimizer and head. This makes the conclusion in §4.2 that 'all multi-source fusion methods outperform their corresponding single-modal baselines' unsupported for this subset of Table 2; the only fair single-source references are the YOLO-family rows, which do not share the fusion methods' architectures. Please add matched single-source versions of the fusion methods under identical training budgets, or re-train the MMRotate baselines for 300 epochs from scratch, and report both.
[§2.4] The semi-supervised annotation procedure contains a circular step: the initial detector is trained and evaluated on the same 5% subset ('with the same data used for both training and evaluation during the initial phase'), and that detector then generates pseudo-labels that are manually corrected and used to expand the labeled set. As described, the initial model's biases can propagate into the majority of the 981,862 instances. Please clarify whether the initial 5% subset was excluded from later training/evaluation rounds, and report agreement statistics between raw pseudo-labels and corrected labels, or a separate validation of label quality on a fully manually annotated sample.

minor comments (6)

[Table 2] The column header for per-class APs is ambiguous: 'Bri. Har. Oil. Ply. Apr. Win.ALL' should be separated clearly and the 'ALL' column defined as the mean over classes.
[§4.2] The sentence stating that the best two results are highlighted 'in Black and mark' is unclear; please specify the intended highlight style (e.g., bold and underline) so the table is interpretable.
[Fig. 9] The label example in Fig. 9 shows duplicate coordinate rows and a confusing arrangement of category IDs; please ensure the example label file exactly matches the described format.
[§9.2] Equation (9) defines L_total = L_reg + L_dfl + L_cls but does not define L_reg, L_dfl, or L_cls; add explicit formulas or references for each loss term.
[Table 1] Table 1 lists M4-SAR resolution as '10M, 60M' while the text uses '10m, 60m'; use consistent units and notation throughout.
[Conclusion] The conclusion claims E2E-OSDet is 'the first fusion framework specifically designed for optical-SAR scenarios,' yet related work cites E2E-MFD [61] and other optical-SAR fusion works; please soften or justify this novelty claim.

Circularity Check

1 steps flagged · score 3.0 of 10

Benchmark labels are partly generated by a detector and then used as ground truth; the claimed 5.7% fusion gain is not directly traceable to Table 2, but the cross-method comparison itself retains independent content.

fitted input called prediction [Section 2.4 (Data Annotation and Label Mapping), Fig. 2(c); evaluation in Table 2.]
"Once manually labeled, a detector [15] is trained using this subset (with the same data used for both training and evaluation during the initial phase). The trained model is then employed to generate pseudo-labels for an additional 5% of the data, which are subsequently refined through manual correction and added to the labeled set."

The annotation detector is fit on the 5% seed subset, and its pseudo-label predictions are iteratively added to the labeled set. After optical-to-SAR label transfer, these labels are the ground truth against which every mAP in Table 2, including E2E-OSDet's 61.4% and the advertised 5.7% fusion gain, is computed. The benchmark's 'ground truth' is therefore partly the output of the same kind of detection model the benchmark is designed to test, and no independent verification or quantification of the manual correction is reported. The loop does not force the relative fusion-vs-single-source ranking, since all methods share the same labels, so this is a partial rather than total circularity.

full rationale

No derivation-level circularity was found: E2E-OSDet is a concrete architecture trained and ablated on M4-SAR, and the fusion-vs-single-source comparison is supported by all fusion methods in Table 2 beating their single-source counterparts, independent of the authors' own detector. The main circular element is in dataset construction (Section 2.4): a detector is trained on a 5% subset and its pseudo-labels become ground truth for the rest of the benchmark, so the label set is partly self-generated; the paper even states the same data are used for training and evaluation in the initial phase. This contaminates absolute mAP values but not the relative ranking across methods, so the score is moderate rather than high. The paper's self-citations (OGSOD-1.0/2.0, MSOD) are used as comparison datasets or related work, not as a load-bearing uniqueness argument. Separately, the abstract's 5.7% fusion improvement is not directly reproducible from Table 2 (best optical gap is 5.1 points, best SAR gap larger), and the weak alignment (SSIM 0.1256, MSE 0.0662, MI 0.1729; geographic-coordinate-only alignment conceded in Section 5.1) is a serious validity risk for the transferred SAR labels, but these are correctness concerns rather than circular derivation steps.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central new artifact is a dataset, not a postulated entity. The method introduces no new physical or mathematical objects; the tunable choices are listed as free parameters. The most consequential assumptions are about label transfer and pseudo-label accuracy, both unmeasured.

free parameters (4)

alpha (handcrafted feature proportion in FAM) = 0.10
Selected in Fig. 5(c) from {0.05,0.10,0.15,0.20,0.25}; 0.10 gives the highest mAP (61.4).
AFM area size k = 2
Selected in Fig. 5(d) from {1,2,3,4,5}; k=2 gives the highest mAP (61.4).
FAM filter type = Grad
Selected in Fig. 5(a) from WST, Canny, Haar, HOG, Grad; Grad gives the highest mAP (61.4).
CMIM scanning mechanism = Bidirectional
Selected in Fig. 5(b) from Bid, Z-order, Zigzag, Hilbert; bidirectional gives the highest mAP (61.4).

assumptions (4)

ad hoc to paper Geographic-coordinate alignment between Sentinel-1 and Sentinel-2 images is sufficient for instance-level label transfer.
Invoked in Sections 2.3-2.4; the paper transfers optical annotations to SAR based on this while reporting SSIM 0.1256, so target positions may shift between modalities.
domain assumption Pseudo-labels from an iteratively retrained detector are accurate enough to label the full dataset without systematic error growth.
Section 2.4 uses this to scale from 5% manual labels to 981,862 instances; no error analysis of pseudo-labels is provided.
ad hoc to paper Cloud-containing optical images inherit labels from their cloud-free counterparts taken within 10 days.
Section 2.4 replaces cloud-free images with cloudy ones after annotation; the optical input may not visibly contain the labeled object.
domain assumption Six coastal-infrastructure categories from four cities are representative enough to benchmark optical-SAR fusion.
Sections 2.2 and 5.1; the paper itself acknowledges limited semantic diversity.

how reviews work

0 comments

Cite this review

Pith. "Pith review of M4-SAR: A Multi-Resolution, Multi-Polarization, Multi-Scene, Multi-Source Dataset and Benchmark for optical-SAR Object Detection." pith.science (2026). https://pith.science/paper/JLIXESUN

@misc{pith2026250510931,
  author       = {Pith},
  title        = {Pith review of: M4-SAR: A Multi-Resolution, Multi-Polarization, Multi-Scene, Multi-Source Dataset and Benchmark for optical-SAR Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JLIXESUN}},
  note         = {Machine review of arXiv:2505.10931}
}

read the original abstract

Single-source remote sensing object detection using optical or SAR images struggles in complex environments. Optical images offer rich textural details but are often affected by low-light, cloud-obscured, or low-resolution conditions, reducing the detection performance. SAR images are robust to weather, but suffer from speckle noise and limited semantic expressiveness. Optical and SAR images provide complementary advantages, and fusing them can significantly improve the detection accuracy. However, progress in this field is hindered by the lack of large-scale, standardized datasets. To address these challenges, we propose a new comprehensive dataset for optical-SAR fusion object detection, named Multi-resolution, Multi-polarization, Multi-scene, Multi-source SAR dataset (M4-SAR). It contains 112,174 instance-level aligned image pairs and nearly one million labeled instances with arbitrary orientations, spanning six key categories. To enable standardized evaluation, we develop a unified benchmarking toolkit that integrates six state-of-the-art multi-source fusion methods. Additionally, we propose E2E-OSDet, a novel end-to-end multi-source fusion detection framework that mitigates cross-domain discrepancies and establishes a robust baseline for future studies. Extensive experiments on M4-SAR demonstrate that fusing optical and SAR data can improve mAP by 5.7\% over single-source inputs, with particularly significant gains in complex environments. The dataset and code are publicly available at https://github.com/wchao0601/M4-SAR.

Figures

Figures reproduced from arXiv: 2505.10931 by the authors.

**Figure 1.** Examples of scenes and six key target categories in the proposed M4-SAR dataset, accompanied by instance size and aspect ratio distributions. multi-source object detection (optical-SAR [22,60]) have garnered significant attention in the remote sensing field. These methods have demonstrated remarkable effectiveness in applications, including disaster monitoring and urban planning. As shown in [PITH_FULL_IMAGE:fig… view at source ↗

**Figure 2.** Pipeline for M4-SAR dataset construction. gaps ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Statistical visualization of category attributes in the proposed M4-SAR dataset. (a) Angle distribution of instances in each category. (b) Percentage of instances per category. (c) Average pixel area of instances per category. large aspect ratios increase the model’s sensitivity to object shapes, thus posing additional challenges to accurate detection. – Angle Diversity Challenge: All target orientation angles are n… view at source ↗

Figures from the paper (11 more)

**Figure 4.** Figure 4: (a) Domain gap between optical and SAR images. (b) Overall framework of the proposed end-to-end optical-SAR fusion object detection (E2E-OSDet). Architectural details of the proposed filter augment module (c), cross-modal Mamba interaction module (d), and area-attentio…

**Figure 5.** Figure 5: Comprehensive performance evaluation under different parameter settings [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

**Figure 6.** Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

**Figure 7.** Figure 7: Qualitative comparison of the proposed E2E-OSDet and six fusion methods. Impact of Filter Augment: To assess the effectiveness of the FAM, we tested a range of conventional feature descriptors. As shown in [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

**Figure 8.** Figure 8: Grad-CAM [36] heatmaps of the proposed E2E-OSDet and six fusion methods. performance without requiring delicate hyperparameter tuning, further underscoring its practicality in real-world deployment. 4.4 Qualitative Analysis As shown in [PITH_FULL_IMAGE:figures/full_f…

**Figure 9.** Figure 9: An example of annotation in our M4-SAR dataset. 7 More details on the M4-SAR 7.1 Dataset and Label Format We introduce M4-SAR, a novel multi-source remote sensing dataset designed to advance research in multi-source object detection. M4-SAR combines highresolution opt…

**Figure 10.** Figure 10: Sample visualization of the M4-SAR dataset. [category,(x1, y1),(x2, y2),(x3, y3),(x4, y4)] (see [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]

**Figure 11.** Figure 11: The scanning mechanisms explored in this work include the Bidirectional Scan, Z-Order, Zigzag, and the 8-direction Hilbert Scanning Method. Cross-modal Mamba Interaction Module: Existing multi-source fusion methods primarily emphasize spatial-level feature integration…

**Figure 12.** Figure 12: Visualization of handcrafted features on SAR images, with SSIM denoting structural similarity. Features are average-pooled and represented as a single channel for visualization [PITH_FULL_IMAGE:figures/full_fig_p029_12.png]

**Figure 13.** Figure 13: Comparison of detection results across different methods (zoom in for detail) [PITH_FULL_IMAGE:figures/full_fig_p030_13.png]

**Figure 14.** Figure 14: Grad-CAM [36] heatmaps of the proposed E2E-OSDet and six fusion methods [PITH_FULL_IMAGE:figures/full_fig_p031_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

SARATR-X-v2: Scale-Aware Structural Pre-Training for SAR Foundation Models
cs.CV 2026-07 conditional novelty 6.0 of 10

A learnable-weighted fusion of six fixed, speckle-robust structural operators as the masked pre-training target transfers better than pixel targets on 10 of 12 SAR benchmarks.

Reference graph

Works this paper leans on

67 extracted references · 60 canonical work pages · cited by 1 Pith paper

[1]

IEEE TPAMI (6), 679–698 (1986)

Canny, J.: A computational approach to edge detection. IEEE TPAMI (6), 679–698 (1986)

work page 1986
[2]

In: CVPR

Cao, Y., Bin, J., Hamari, J., Blasch, E., Liu, Z.: Multimodal object detection by channel switching and spatial attention. In: CVPR. pp. 403–411 (2023)

work page 2023
[3]

IEEE TGRS60, 1–11 (2022)

Cheng, G., Wang, J., Li, K., Xie, X., Lang, C., Yao, Y., Han, J.: Anchor-free oriented proposal generator for object detection. IEEE TGRS60, 1–11 (2022)

work page 2022
[4]

In: CVPR

Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: CVPR. vol. 1, pp. 886–893 (2005)

work page 2005
[5]

IEEE TPAMI44(11), 7778–7796 (2021)

Ding, J., Xue, N., Xia, G.S., Bai, X., Yang, W., Yang, M.Y., Belongie, S., Luo, J., Datcu, M., Pelillo, M., et al.: Object detection in aerial images: A large-scale benchmark and challenges. IEEE TPAMI44(11), 7778–7796 (2021)

work page 2021
[6]

IEEE TMM (2025)

Dong, W., Zhu, H., Lin, S., Luo, X., Shen, Y., Guo, G., Zhang, B.: Fusion-mamba for cross-modality object detection. IEEE TMM (2025)

work page 2025
[7]

RSE120, 25–36 (2012)

Drusch, M., Del Bello, U., Carlier, S., Colin, O., Fernandez, V., Gascon, F., Ho- ersch, B., Isola, C., Laberinti, P., Martimort, P., et al.: Sentinel-2: Esa’s optical high-resolution mission for gmes operational services. RSE120, 25–36 (2012)

work page 2012
[8]

IEEE TCSVT 33(9), 4754–4768 (2023)

Geng, P., Lu, X., Hu, C., Liu, H., Lyu, L.: Focusing fine-grained action by self- attention-enhanced graph neural networks with contrastive learning. IEEE TCSVT 33(9), 4754–4768 (2023)

work page 2023

Show all 67 references

[9]

In: ICCV

Girshick, R.: Fast r-cnn. In: ICCV. pp. 1440–1448 (2015)

2015
[10]

In: COLM (2024)

Gu, A., Dao, T.: Mamba: Linear-time sequence modeling with selective state spaces. In: COLM (2024)

2024
[11]

NeurIPS37, 71162–71187 (2024)

He, H., Bai, Y., Zhang, J., He, Q., Chen, H., Gan, Z., Wang, C., Li, X., Tian, G., Xie, L.: Mambaad: Exploring state space models for multi-class unsupervised anomaly detection. NeurIPS37, 71162–71187 (2024)

2024
[12]

In: ICCV

He, K., Gkioxari, G., Dollár, P., Girshick, R.: Mask r-cnn. In: ICCV. pp. 2961–2969 (2017)

2017
[13]

In: ACM MM

He, X., Tang, C., Zou, X., Zhang, W.: Multispectral object detection via cross- modal conflict-aware learning. In: ACM MM. pp. 1465–1474 (2023)

2023
[14]

In: AAAI

Hu,X.,Tai,Y.,Zhao,X.,Zhao,C.,Zhang,Z.,Li,J.,Zhong,B.,Yang,J.:Exploiting multimodal spatial-temporal patterns for video object tracking. In: AAAI. vol. 39, pp. 3581–3589 (2025)

2025
[15]

com/ultralytics/ultralytics

Jocher, G., Qiu, J., Chaurasia, A.: Ultralytics YOLO (Jan 2023),https://github. com/ultralytics/ultralytics

2023
[16]

arXiv preprint arXiv:2209.02976 (2022)

Li, C., Li, L., Jiang, H., Weng, K., Geng, Y., Li, L., Ke, Z., Li, Q., Cheng, M., Nie, W., et al.: Yolov6: A single-stage object detection framework for industrial applications. arXiv preprint arXiv:2209.02976 (2022)

2022 arXiv
[17]

IEEE TIP (2024)

Li, J., Wang, X., Zhao, H., Zhong, Y.: Learning a cross-modality anomaly detector for remote sensing imagery. IEEE TIP (2024)

2024
[18]

In: CVPR

Li, K., Wang, D., Hu, Z., Zhu, W., Li, S., Wang, Q.: Unleashing channel poten- tial: Space-frequency selection convolution for sar object detection. In: CVPR. pp. 17323–17332 (2024)

2024
[19]

IEEE TIP (2025)

Li, W., Yang, W., Hou, Y., Liu, L., Liu, Y., Li, X.: Saratr-x: Towards building a foundation model for sar target recognition. IEEE TIP (2025)

2025
[20]

ISPRS218, 326–338 (2024) 16 C

Li, W., Yang, W., Liu, T., Hou, Y., Li, Y., Liu, Z., Liu, Y., Liu, L.: Predicting gradi- ent is better: Exploring self-supervised learning for sar atr with a joint-embedding predictive architecture. ISPRS218, 326–338 (2024) 16 C. Wang et al

2024
[21]

In: CVPR

Li, W., Chen, Y., Hu, K., Zhu, J.: Oriented reppoints for aerial object detection. In: CVPR. pp. 1829–1838 (2022)

2022
[22]

ISPRS179, 14–34 (2021)

Li, X., Du, Z., Huang, Y., Tan, Z.: A deep translation (gan) based change detection network for optical and sar remote sensing images. ISPRS179, 14–34 (2021)

2021
[23]

Li, Y., Li, X., Dai, Y., Hou, Q., Liu, L., Liu, Y., Cheng, M.M., Yang, J.: Lsknet: A foundation lightweight backbone for remote sensing. IJCV pp. 1–22 (2024)

2024
[24]

NeurIPS37, 128430–128461 (2024)

Li, Y., Li, X., Li, W., Hou, Q., Liu, L., Cheng, M.M., Yang, J.: Sardet-100k: Towards open-source benchmark and toolkit for large-scale sar object detection. NeurIPS37, 128430–128461 (2024)

2024
[25]

RS15(11), 2825 (2023)

Lin, X., Zhang, B., Wu, F., Wang, C., Yang, Y., Chen, H.: Sived: A sar image dataset for vehicle detection based on rotatable bounding box. RS15(11), 2825 (2023)

2023
[26]

arXiv preprint arXiv:2512.23273 (2025)

Lin, X., Peng, J., Gan, Z., Zhu, J., Liu, J.: Yolo-master: Moe-accelerated with specialized transformers for enhanced real-time detection. arXiv preprint arXiv:2512.23273 (2025)

2025 arXiv
[27]

Information Fusion125, 103414 (2026)

Liu, C., Ma, X., Yang, X., Zhang, Y., Dong, Y.: Como: Cross-mamba interaction and offset-guided fusion for multimodal object detection. Information Fusion125, 103414 (2026)

2026
[28]

In: ICPRAM

Liu, Z., Yuan, L., Weng, L., Yang, Y.: A high resolution optical satellite image dataset for ship recognition and some new baselines. In: ICPRAM. vol. 2, pp. 324–331 (2017)

2017
[29]

arXiv preprint arXiv:2106.06072 (2021)

Llerena, J.M., Zeni, L.F., Kristen, L.N., Jung, C.: Gaussian bounding boxes and probabilistic intersection-over-union for object detection. arXiv preprint arXiv:2106.06072 (2021)

2021 arXiv
[30]

arXiv preprint arXiv:2501.10040 (2025)

Lu,W.,Chen,S.B.,Ding,C.H., Tang,J.,Luo,B.: Lwganet: Alightweightgroupat- tention backbone for remote sensing visual tasks. arXiv preprint arXiv:2501.10040 (2025)

2025
[31]

ISPRS231, 431–442 (2026)

Lu, W., Li, H.D., Wang, C., Chen, S.B., Ding, C.H., Tang, J., Luo, B.: Unravelnet: A backbone for enhanced multi-scale and low-quality feature extraction in remote sensing object detection. ISPRS231, 431–442 (2026)

2026
[32]

arXiv preprint arXiv:2212.07784 (2022)

Lyu, C., Zhang, W., Huang, H., Zhou, Y., Wang, Y., Liu, Y., Zhang, S., Chen, K.: Rtmdet: An empirical study of designing real-time object detectors. arXiv preprint arXiv:2212.07784 (2022)

2022 arXiv
[33]

CPAM65(10), 1331–1398 (2012)

Mallat, S.: Group invariant scattering. CPAM65(10), 1331–1398 (2012)

2012
[34]

MMRotate Contributors: OpenMMLab rotated object detection toolbox and benchmark (Feb 2022),https://github.com/open-mmlab/mmrotate

2022
[35]

PR130, 108786 (2022)

Qingyun, F., Zhaokui, W.: Cross-modality attentive feature fusion for object de- tection in multispectral remote sensing imagery. PR130, 108786 (2022)

2022
[36]

In: ICCV

Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D.: Grad- cam: Visual explanations from deep networks via gradient-based localization. In: ICCV. pp. 618–626 (2017)

2017
[37]

PR145, 109913 (2024)

Shen, J., Chen, Y., Liu, Y., Zuo, X., Fan, H., Yang, W.: Icafusion: Iterative cross- attention guided feature fusion for multispectral object detection. PR145, 109913 (2024)

2024
[38]

IEEE TIV (2024)

Song, K., Xue, X., Wen, H., Ji, Y., Yan, Y., Meng, Q.: Misaligned visible-thermal object detection: a drone-based benchmark and baseline. IEEE TIV (2024)

2024
[39]

AAAI (2025)

Sun, Y., Wang, C., Yang, J., Luo, L.: Small but mighty: Dynamic wavelet expert- guided fine-tuning of large-scale models for optical remote sensing object segmen- tation. AAAI (2025)

2025
[40]

IEEE TGRS62, 1–14 (2024) Abbreviated paper title 17

Sun, Y., Yang, J., Luo, L.: United domain cognition network for salient object detection in optical remote sensing images. IEEE TGRS62, 1–14 (2024) Abbreviated paper title 17

2024
[41]

arXiv preprint arXiv:2502.12524 (2025)

Tian, Y., Ye, Q., Doermann, D.: Yolov12: Attention-centric real-time object detec- tors. arXiv preprint arXiv:2502.12524 (2025)

2025 arXiv
[42]

In: ICCV

Tian, Z., Shen, C., Chen, H., He, T.: Fcos: Fully convolutional one-stage object detection. In: ICCV. pp. 9627–9636 (2019)

2019
[43]

RSE120, 9–24 (2012)

Torres, R., Snoeij, P., Geudtner, D., Bibby, D., Davidson, M., Attema, E., Potin, P., Rommen, B., Floury, N., Brown, M., et al.: Gmes sentinel-1 mission. RSE120, 9–24 (2012)

2012
[44]

In: CVPR

Viola, P., Jones, M.: Rapid object detection using a boosted cascade of simple features. In: CVPR. vol. 1, pp. I–I (2001)

2001
[45]

NeurIPS37, 107984–108011 (2024)

Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., Han, J., et al.: Yolov10: Real-time end-to-end object detection. NeurIPS37, 107984–108011 (2024)

2024
[46]

IEEE TGRS 63, 1–13 (2025)

Wang, C., Fang, W., Li, X., Yang, J., Luo, L.: Msod: A large-scale multiscene dataset and a novel diagonal-geometry loss for sar object detection. IEEE TGRS 63, 1–13 (2025)

2025
[47]

In: ICASSP

Wang, C., Luo, L., Fang, W., Yang, J.: Cross-modal gaussian localization distil- lation for optical information guided sar object detection. In: ICASSP. pp. 1–5 (2025)

2025
[48]

IEEE TGRS61, 1–14 (2023)

Wang, C., Ruan, R., Zhao, Z., Li, C., Tang, J.: Category-oriented localization distillation for sar object detection and a unified benchmark. IEEE TGRS61, 1–14 (2023)

2023
[49]

IEEE TCSVT (2026)

Wang, C., Sun, Y., Yang, J., Luo, L.: Localized background-aware generative dis- tillation for enhanced remote sensing object detection. IEEE TCSVT (2026)

2026
[50]

In: ECCV

Wang, C.Y., Yeh, I.H., Mark Liao, H.Y.: Yolov9: Learning what you want to learn using programmable gradient information. In: ECCV. pp. 1–21 (2024)

2024
[51]

IEEE TIP32, 565–579 (2022)

Wang, J., Cui, Z., Jiang, T., Cao, C., Cao, Z.: Lightweight deep neural networks for ship target detection in sar imagery. IEEE TIP32, 565–579 (2022)

2022
[52]

RS11(7), 765 (2019)

Wang, Y., Wang, C., Zhang, H., Dong, Y., Wei, S.: A sar dataset of ship detection for deep learning under complex backgrounds. RS11(7), 765 (2019)

2019
[53]

IEEE Access8, 120234–120254 (2020)

Wei, S., Zeng, X., Qu, Q., Wang, M., Su, H., Shi, J.: Hrsid: A high-resolution sar images dataset for ship detection and instance segmentation. IEEE Access8, 120234–120254 (2020)

2020
[54]

IEEE TGRS (2024)

Wu, Y., Suo, Y., Meng, Q., Dai, W., Miao, T., Zhao, W., Yan, Z., Diao, W., Xie, G., Ke, Q., et al.: Fair-csar: A benchmark dataset for fine-grained object detection and recognition based on single look complex sar images. IEEE TGRS (2024)

2024
[55]

RS14(6), 1488 (2022)

Xia, R., Chen, J., Huang, Z., Wan, H., Wu, B., Sun, L., Yao, B., Xiang, H., Xing, M.: Crtranssar: A visual transformer based on contextual joint represen- tation learning for sar ship detection. RS14(6), 1488 (2022)

2022
[56]

In: CVPR

Yu, Y., Da, F.: Phase-shifting coder: Predicting accurate orientation in oriented object detection. In: CVPR. pp. 13354–13363 (2023)

2023
[57]

In: ECCV

Yuan, M., Wang, Y., Wei, X.: Translation, scale and rotation: cross-modal align- ment meets rgb-infrared vehicle detection. In: ECCV. pp. 509–525 (2022)

2022
[58]

IEEE TCSVT (2024)

Zeng, Y., Liang, T., Jin, Y., Li, Y.: Mmi-det: Exploring multi-modal integration for visible and infrared object detection. IEEE TCSVT (2024)

2024
[59]

Neurocomputing 631, 129718 (2025)

Zhan, Y., Zeng, Z., Liu, H., Tan, X., Tian, Y.: Mambasod: Dual mamba-driven cross-modal fusion network for rgb-d salient object detection. Neurocomputing 631, 129718 (2025)

2025
[60]

IJAEOG109, 102769 (2022) 18 C

Zhang, C., Feng, Y., Hu, L., Tapete, D., Pan, L., Liang, Z., Cigna, F., Yue, P.: A domain adaptation neural network for change detection with heterogeneous optical and sar remote sensing images. IJAEOG109, 102769 (2022) 18 C. Wang et al

2022
[61]

In: NeurIPS (2024)

Zhang, J., Cao, M., Yang, X., Xie, W., Lei, J., Li, D., Yang, G., Huang, W., Li, Y.: E2e-mfd: Towards end-to-end synchronous multimodal fusion detection. In: NeurIPS (2024)

2024
[62]

IEEE JSTARS15, 3365–3375 (2022)

Zhang, P., Xu, H., Tian, T., Gao, P., Li, L., Zhao, T., Zhang, N., Tian, J.: Se- fepnet: Scale expansion and feature enhancement pyramid network for sar aircraft detection with small sample dataset. IEEE JSTARS15, 3365–3375 (2022)

2022
[63]

In: CVPR

Zhang, S., Chi, C., Yao, Y., Lei, Z., Li, S.Z.: Bridging the gap between anchor- based and anchor-free detection via adaptive training sample selection. In: CVPR. pp. 9759–9768 (2020)

2020
[64]

RS13(18), 3690 (2021)

Zhang, T., Zhang, X., Li, J., Xu, X., Wang, B., Zhan, X., Xu, Y., Ke, X., Zeng, T., Su, H., et al.: Sar ship detection dataset (ssdd): Official release and comprehensive data analysis. RS13(18), 3690 (2021)

2021
[65]

ISPRS 231, 151–166 (2026)

Zhang, W., Zhang, X., Xu, X., Wei, S., Shi, J., Wang, Y., Zeng, T.: Mhfnet: Mul- timodal hybrid fusion framework for misaligned sar-optical ship detection. ISPRS 231, 151–166 (2026)

2026
[66]

In: CVPR

Zhang, X., Yang, X., Li, Y., Yang, J., Cheng, M.M., Li, X.: Rsar: Restricted state angle resolver and rotated sar benchmark. In: CVPR. pp. 7416–7426 (2025)

2025
[67]

Zou, Z., Shi, Z.: Random access memories: A new paradigm for target detection in high resolution aerial remote sensing images. IEEE TIP27(3), 1100–1111 (2017) Abbreviated paper title 19 Supplementary Material: #146 This appendix is organized as follows: – Appendix 7provides co...

2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

IEEE TPAMI (6), 679–698 (1986)

Canny, J.: A computational approach to edge detection. IEEE TPAMI (6), 679–698 (1986)

work page 1986

[2] [2]

In: CVPR

Cao, Y., Bin, J., Hamari, J., Blasch, E., Liu, Z.: Multimodal object detection by channel switching and spatial attention. In: CVPR. pp. 403–411 (2023)

work page 2023

[3] [3]

IEEE TGRS60, 1–11 (2022)

Cheng, G., Wang, J., Li, K., Xie, X., Lang, C., Yao, Y., Han, J.: Anchor-free oriented proposal generator for object detection. IEEE TGRS60, 1–11 (2022)

work page 2022

[4] [4]

In: CVPR

Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: CVPR. vol. 1, pp. 886–893 (2005)

work page 2005

[5] [5]

IEEE TPAMI44(11), 7778–7796 (2021)

Ding, J., Xue, N., Xia, G.S., Bai, X., Yang, W., Yang, M.Y., Belongie, S., Luo, J., Datcu, M., Pelillo, M., et al.: Object detection in aerial images: A large-scale benchmark and challenges. IEEE TPAMI44(11), 7778–7796 (2021)

work page 2021

[6] [6]

IEEE TMM (2025)

Dong, W., Zhu, H., Lin, S., Luo, X., Shen, Y., Guo, G., Zhang, B.: Fusion-mamba for cross-modality object detection. IEEE TMM (2025)

work page 2025

[7] [7]

RSE120, 25–36 (2012)

Drusch, M., Del Bello, U., Carlier, S., Colin, O., Fernandez, V., Gascon, F., Ho- ersch, B., Isola, C., Laberinti, P., Martimort, P., et al.: Sentinel-2: Esa’s optical high-resolution mission for gmes operational services. RSE120, 25–36 (2012)

work page 2012

[8] [8]

IEEE TCSVT 33(9), 4754–4768 (2023)

Geng, P., Lu, X., Hu, C., Liu, H., Lyu, L.: Focusing fine-grained action by self- attention-enhanced graph neural networks with contrastive learning. IEEE TCSVT 33(9), 4754–4768 (2023)

work page 2023