Pith. sign in

REVIEW 4 major objections 3 minor 44 references

Bridging Severe Cross-Modal Misalignment: End-to-End Visible-Infrared Object Detection via Explicit Feature-Domain Affine Registration

T0 review · 4 major / 3 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Explicitly registering infrared features to the visible reference before fusion lifts severe-misalignment visible-infrared detection to 69.7% mAP50 on a new benchmark.

desk verdict A coherent system paper whose headline SOTA claim is undermined by privileged geometric supervision that baselines don't receive; the idea is worth scrutiny, but the current evidence only shows the method works when it sees the warp parameters. read the letter →

arxiv 2608.10680 v1 pith:TQZNSGXC submitted 2026-08-11 cs.CV

Qi Ming , Yuyang Wang , Mingjing Zhao , Yifan Xiao , Zhixin Guo , Zhiqiang Zhou , Peng Sun , Juan Fang
show 2 more authors
Fuqiang Yang Xudong Zhao
This is my paper · ORCID
classification cs.CV
keywords visible-infraredobjectdetectionorientedcross-modalmisalignmentaffinefeatureregistrationillumination-guidedfusionmultimodalbenchmarkthermalinfrareddroneimagery
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

The paper addresses visible-infrared object detection when the two images are severely misaligned — offsets of tens to hundreds of pixels, plus rotation and scale differences — rather than the weakly aligned pairs most prior work assumes. It argues that implicit feature adaptation is insufficient at this scale and that explicit, image-level affine registration in the feature domain, done end-to-end with detection, is the missing piece. To make the case, the authors build DVMA, a benchmark derived from DroneVehicle by applying controlled rotation, anisotropic scaling, and translation to infrared images, and report that their JFRDet reaches 69.7% mAP50, outperforming the best compared method (DMM + S2A-Net at 66.7%) by about 3 points. If true, the result indicates that global geometric correction is a practical first-class step for misaligned multimodal detection, not merely a preprocessing nicety.

What carries the argument

The load-bearing object is the image-level affine transformation $T_{\mathrm{ir}\to\mathrm{rgb}}$ estimated by Cross-Modal Affine Alignment (CMAA): a $2\times3$ matrix mapping infrared coordinates to visible coordinates, fitted by confidence-weighted least squares from coarse-to-fine bidirectional correspondence matches and then applied to every level of the infrared feature pyramid through bilinear warping. It converts cross-modal fusion from an operation on misaligned feature maps into one on a common geometric reference, which is what lets the detector exploit complementary cues despite large offsets. The two supporting mechanisms are Illumination-Guided Complementary Fusion (IGCF), which scales a discrepancy-based spatial gate by a darkness factor computed from mean visible intensity, and Alignment Quality-Consistency Gating (AQCG), which gates the detection loss by an exponential moving average of alignment loss and cosine similarity of alignment/detection gradients.

What would settle it

Run JFRDet on a real dual-sensor rig with rigidly offset cameras that produce measured large misalignments involving parallax, and compare mAP50 against DMM + S2A-Net: the affine-registration claim would be falsified if the gap shrinks to noise or if the estimated transform leaves visible feature displacement.

Watch

Extended reading notes

Core claim

JFRDet is an end-to-end visible-infrared oriented object detector built around a Cross-Modal Affine Alignment (CMAA) module. CMAA first finds coarse bidirectional correspondences between RGB and infrared features at 1/16 resolution, refines them locally, and then estimates a single affine transformation $T_{\mathrm{ir}\to\mathrm{rgb}}\in\mathbb{R}^{2\times3}$ by confidence-weighted least squares. That transform is rescaled to each feature level and applied by inverse bilinear grid sampling, so every infrared pyramid feature is warped to the visible reference before fusion. Fusion is then gated by an illumination estimate, and an Alignment Quality-Consistency Gating (AQCG) mechanism reweights detection loss according to how reliable and gradient-consistent the alignment is. On the new DVMA benchmark, built from DroneVehicle pairs with rotations of 15–30 degrees, scaling factors of 0.90–1.10, and translations producing overlaps of 0.70–0.80 IoU, JFRDet obtains 69.7% mAP50 and 36.1% mAP50:95, the best numbers reported in the paper, ahead of the strongest baseline DMM + S2A-Net at 66.7% and 35.1%.

Load-bearing premise

The whole method assumes that the spatial mismatch between a visible-infrared image pair can be captured by one global 2D affine transformation, and that the synthetic affine distortions used to build the test benchmark behave like the severe misalignment found in real scenes.

Editorial extensions

If this is right

  • On DVMA, the full JFRDet reaches 69.7% mAP50, which is 3.0 points above the best compared method (DMM + S2A-Net at 66.7%).
  • The component ablation attributes the gain to each mechanism: affine alignment alone moves the no-alignment fusion baseline from 66.0% to 67.3%, and adding illumination-guided fusion and gating brings it to 69.7%.
  • The comparison also shows that infrared-only S2A-Net (61.6% mAP50) beats several RGB-IR fusion methods on DVMA, so unaligned fusion can be worse than a single modality; explicit alignment is what makes the extra modality pay off.
  • Because DVMA records exact transformation parameters, it supplies per-pixel correspondence supervision at every feature resolution, enabling direct supervised training of cross-modal registration rather than relying on implicit adaptation.

Reading between the lines

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

  • Editorial inference: If real misalignment includes parallax, independent object motion, or lens distortion, one global affine transform cannot represent it; a natural next test is replacing the image-level fit with piecewise-affine or dense-flow refinement after an affine warm start.
  • Editorial inference: The DVMA benchmark's synthetic perturbations and known ground truth could serve as a pretraining or calibration stage for real misaligned pairs, where true correspondences are unavailable.
  • Editorial inference: The illumination gate in IGCF uses only global mean visible intensity; extending it to a local or learned illumination estimate is a direct, testable variant that could matter for scenes with mixed shadows or strong light sources.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 3 minor

Summary. The paper proposes JFRDet, an end-to-end visible-infrared oriented object detector targeting severe cross-modal spatial misalignment. The method has three main components: a Cross-Modal Affine Alignment (CMAA) module that estimates an image-level affine transform from visible to infrared and warps multi-level infrared features; an Illumination-Guided Complementary Fusion (IGCF) module that adaptively suppresses unreliable visible cues under poor illumination; and an Alignment Quality-Consistency Gating (AQCG) strategy that reweights detection supervision based on alignment quality and gradient consistency. The authors also construct DVMA, a synthetic benchmark derived from DroneVehicle by applying random global affine transforms to the infrared image and recording the warp parameters to generate dense correspondence labels. On DVMA, JFRDet reports 69.7% mAP50 and 36.1% mAP50:95, outperforming DMM+S2A-Net by 3.0 mAP50, with component ablations from 66.0 to 69.7 mAP50.

Significance. Explicit feature-domain affine registration for severely misaligned visible-infrared detection is a worthwhile and underexplored problem, and the paper contains a clean formulation of coarse-to-fine correspondence, confidence-weighted affine fitting, and differentiable warping. The proposed DVMA benchmark could be a useful testbed if its limitations are clearly stated. However, the central SOTA claim is currently supported only by experiments on a synthetic benchmark whose generative process is exactly the global affine model assumed by CMAA, and the comparison is potentially confounded by privileged geometric supervision that the baselines do not receive. The mAP50:95 margin over the best baseline is only 1.0 point. These issues make the quantitative claims provisional rather than established.

major comments (4)
  1. [Sec. 3.1 / Sec. 3.5 / Tables 2-3] The reported SOTA gain is confounded by privileged affine-correspondence supervision. DVMA records all transformation parameters, and Sec. 3.1 states that exact cross-modal correspondences are generated at each feature resolution and that correspondence supervision is provided in both directions; these labels are used by CMAA's alignment loss L_align. None of the compared baselines (S2A-Net, DMM, C2Former, COMO) receive such geometric supervision. Therefore the +3.0 mAP50 over DMM+S2A-Net in Table 2, and the CMAA ablation in Table 3, do not isolate the contribution of the architecture: the gain could come from seeing ground-truth warp parameters. Please add a control experiment in which a comparable baseline is given the same privileged information (for example, pre-warping infrared features with the recorded affine transform, or training a simple alignment module with the same offset-field supervision), and/or train CMAA without correspondence supervision to measure the architecture-only gain.
  2. [Sec. 3.1 / Sec. 3.3.3 / Abstract] The external validity of the benchmark is not established. DVMA is constructed by applying a single global 2D affine transform to the infrared image, and CMAA estimates exactly such an image-level affine transform. The test distribution is therefore generated from the same model class the method assumes. This does not support the abstract's broad claim of 'bridging severe cross-modal misalignment' in real scenes, where parallax, independent object motion, and non-affine distortion occur. Please evaluate on a real misaligned dataset or augment DVMA with non-affine and object-level perturbations, and temper the claims to the global-affine setting.
  3. [Figure 3] The central architecture figure is inconsistent with the described method. Figure 3 contains elements such as 'Hyperspectral image', 'SGMCT framework', 'Spectral-Guided Fusion Module (SGFM)', and 'RS2Conv' that are not part of JFRDet and appear to come from an unrelated manuscript on hyperspectral image classification. This must be corrected with an actual diagram of CMAA, IGCF, and AQCG, and the provenance of the current figure must be clarified. As it stands, the visual evidence for the architecture cannot be trusted.
  4. [Sec. 4.3 / Tables 2-3] All results are single-run, and code and dataset are not released. With a 3.0-point mAP50 margin and no variance estimate, the SOTA claim is not statistically grounded; additionally, the mAP50:95 margin over the best baseline is only 1.0 point (36.1 vs 35.1), so the improvement is much smaller on the stricter localization metric. Please report multiple seeds with mean and standard deviation, and release code and dataset or specify a detailed evaluation protocol.
minor comments (3)
  1. [Sec. 3.3.1] Please define 'Coarse Matching Matrices' before using the term; the current text jumps from the similarity matrix S to the bidirectional confidence matrices without specifying how the coarse matching matrices are formed.
  2. [Figure 2 / Sec. 3.1] Figure 2 shows both 'Transformed RGB' and 'Transformed IR', while the text says only the infrared image undergoes a compound transformation and the visible image is center-cropped. Please clarify whether the RGB branch is also resized or transformed.
  3. [Sec. 4.2] Please state how oriented ground-truth boxes are handled when images are resized to 480x384, and whether the recorded affine parameters are used to transform the annotations during DVMA construction; this affects the interpretation of the reported AP values.

Circularity Check

1 steps flagged · score 4.0 of 10

DVMA is co-designed with JFRDet's affine mechanism: the benchmark is generated by recorded affine warps and CMAA is trained on exact correspondences derived from those warps, so the reported SOTA comparison is confounded by privileged supervision.

  1. self definitional [Section 3.1 (DVMA construction), Section 3.3.3 (Feature Affine Registration), Section 3.5 (L_align); evaluation in Table 2]
    "Since all transformation parameters are recorded, exact cross-modal correspondences can be generated automatically at each feature resolution. ... DVMA provides explicit alignment annotations for paired images, enabling supervised cross-modal registration during training."

    DVMA's 'compound transformation composed of rotation, anisotropic scaling, and translation' is precisely the image-level affine model that CMAA is designed to estimate, as stated in Section 3.3.3: 'CMAA estimates an image-level affine transformation from the infrared modality to the visible reference.' The recorded parameters are used to synthesize dense correspondence and offset targets consumed by the alignment loss, so the central alignment module is trained under the same generative model that defines the benchmark's test distribution. The compared baselines in Table 2 do not receive this alignment ground truth, so the reported 69.7% versus 66.7% mAP50 gain conflates the architecture's contribution with the effect of privileged affine supervision.

full rationale

The SOTA claim is an empirical result, not a theorem, so no derivation is circular in the equation sense. However, the evaluation is co-designed with the method: DVMA is synthesized by applying recorded rotation, anisotropic-scaling, and translation warps to IR images, and those same recordings generate dense correspondence and offset-field targets for CMAA's alignment loss. The test distribution is therefore exactly the affine model class JFRDet is engineered to invert, while the compared methods never receive the alignment annotations. This makes the +3.0 mAP50 gain over DMM+S2A-Net a confounded comparison rather than an independent demonstration of the architecture's value, and it limits external validity to scenes whose misalignment is global-affine; parallax or independent object motion are outside the benchmark. Because the method still must learn and the numbers are measured, this is a partial circularity of benchmark construction, not a derivation that reduces to its inputs by construction; score 4.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. Its free parameters are largely unstated hyperparameters for matching, illumination, and gating. The key axioms are domain assumptions: misalignment is globally affine, mean intensity measures illumination reliability, IR labels remain valid after perturbation, and gradient consistency proxies objective agreement. These assumptions are load-bearing and untested against non-affine misalignment.

free parameters (9)
  • Coarse matching confidence threshold theta_c = not reported
    Filters low-confidence coarse correspondences in CMAA; chosen by hand, value absent from the paper.
  • Fine matching confidence threshold theta_f = not reported
    Filters local window matches; chosen by hand, value absent.
  • Illumination threshold eta_0 = not reported
    Defines darkness factor d(eta) in IGCF; chosen by hand.
  • Similarity temperature tau = not reported
    Scales the similarity matrix in coarse and fine matching; controls softmax sharpness.
  • AQCG quality threshold theta = 0.50 (best in Table 5)
    Tuned by sensitivity analysis; best value reported, but the selection set is not specified.
  • AQCG balance factor lambda_c = 0.30 (best in Table 5)
    Tuned balance between alignment quality and gradient consistency.
  • AQCG transition sharpness kappa = not reported
    Controls tanh sharpness of the gating signal.
  • AQCG EMA momentum beta and reference loss L_ref = not reported
    Used to compute alignment quality Q_t; neither value is specified.
  • Attention depth N in coarse matching = not reported
    Number of self-attention and cross-attention layers in the coarse matching stack is not specified.
assumptions (4)
  • domain assumption Cross-modal misalignment can be represented by a single global 2D affine transform.
    CMAA fits one image-level affine transform T_ir->rgb (Section 3.3.3), and DVMA is constructed by applying exactly such transforms to the IR image (Section 3.1). If real misalignment is non-affine, the central mechanism fails.
  • domain assumption Infrared labels remain valid and complete after the DVMA affine perturbation.
    Section 4.2 uses ground-truth annotations from the infrared modality as training labels, but the paper does not explicitly describe how detection boxes are transformed along with the IR image during DVMA construction.
  • domain assumption Mean pixel intensity of the visible image is a sufficient proxy for illumination and RGB reliability.
    Section 3.4 computes the illumination score eta as normalized mean intensity and derives the darkness factor from it; this ignores local lighting, glare, and sensor gain.
  • domain assumption Cosine similarity between alignment and detection gradients with respect to T is a valid proxy for joint optimization stability.
    Section 3.5 uses C_t = cos_sim(dL_align/dT, dL_det/dT) to gate detection supervision; the paper does not validate that this correlation actually indicates a stable joint optimum.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Severe Cross-Modal Misalignment: End-to-End Visible-Infrared Object Detection via Explicit Feature-Domain Affine Registration." pith.science (2026). https://pith.science/paper/TQZNSGXC

@misc{pith2026260810680,
  author       = {Pith},
  title        = {Pith review of: Bridging Severe Cross-Modal Misalignment: End-to-End Visible-Infrared Object Detection via Explicit Feature-Domain Affine Registration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TQZNSGXC}},
  note         = {Machine review of arXiv:2608.10680}
}
abstract

Visible-infrared object detection relies on complementary RGB and thermal cues, but its performance is often degraded by cross-modal spatial misalignment. Most existing methods rely on implicit feature adaptation to handle weakly misaligned scenarios, while large-offset geometric discrepancies remain insufficiently addressed. In this paper, we propose a Joint Feature-domain Registration and Detection network (JFRDet), an end-to-end visible-infrared oriented object detector tailored for severely cross-modal geometric discrepancies. JFRDet introduces a Cross-Modal Affine Alignment (CMAA) module to estimate an image-level affine transformation for explicit multi-level feature alignment. Note that illumination changes directly affect the reliability of RGB cues, an Illumination-Guided Complementary Fusion (IGCF) module adaptively exploits modality reliability under varying illumination conditions for cross-modal fusion. Then, an Alignment Quality-Consistency Gating (AQCG) strategy stabilizes joint optimization by modulating detection supervision according to alignment reliability and gradient consistency. We further construct DroneVehicle Misaligned (DVMA), a benchmark for evaluating visible-infrared oriented object detection under severe cross-modal geometric misalignment. The proposed JFRDet achieves 69.7\% $\mathrm{mAP}_{50}$ on DVMA, which represents state-of-the-art (SOTA) performance. The code and dataset will be available on GitHub.

Figures

Figures reproduced from arXiv: 2608.10680 by the authors.

Figure 1
Figure 1. Comparison of different multimodal detection paradigms. (a) Spatially aligned multimodal detection assumes well-registered [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Construction pipeline of the DVMA benchmark. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Network architecture of the proposed JFRDet. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on DVMA at a confidence [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualization of feature alignment by CMAA. From [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 39 canonical work pages

  1. [1]

    Weakly misalignment-free adaptive feature alignment for uavs- based multimodal object detection

    Chen Chen, Jiahao Qi, Xingyue Liu, Kangcheng Bin, Ruigang Fu, Xikun Hu, and Ping Zhong. Weakly misalignment-free adaptive feature alignment for uavs- based multimodal object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 26836–26845, June 2024. 1

    work page 2024

  2. [2]

    Multi-layer fu- sion techniques using a cnn for multispectral pedestrian de- tection.IET Computer Vision, 12(8):1179–1187, 2018

    Yunfan Chen, Han Xie, and Hyunchul Shin. Multi-layer fu- sion techniques using a cnn for multispectral pedestrian de- tection.IET Computer Vision, 12(8):1179–1187, 2018. 3

    work page 2018

  3. [3]

    Learning roi transformer for oriented object detection in aerial images

    Jian Ding, Nan Xue, Yang Long, Gui-Song Xia, and Qikai Lu. Learning roi transformer for oriented object detection in aerial images. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2849– 2858, 2019. 2

    work page 2019

  4. [4]

    Cf-deformable detr: An end-to-end alignment-free model for weakly aligned visible-infrared ob- ject detection

    Haolong Fu, Jin Yuan, Guojin Zhong, Xuan He, Jiacheng Lin, and Zhiyong Li. Cf-deformable detr: An end-to-end alignment-free model for weakly aligned visible-infrared ob- ject detection. InIJCAI, pages 758–766, 2024. 1

    work page 2024

  5. [5]

    Damsdet: Dynamic adaptive multispectral de- tection transformer with competitive query selection and adaptive feature fusion

    Junjie Guo, Chenqiang Gao, Fangcen Liu, Deyu Meng, and Xinbo Gao. Damsdet: Dynamic adaptive multispectral de- tection transformer with competitive query selection and adaptive feature fusion. InEuropean Conference on Com- puter Vision, pages 464–481. Springer, 2024. 1

    work page 2024

  6. [6]

    Beyond bounding-box: Convex- hull feature adaptation for oriented and densely packed ob- ject detection

    Zonghao Guo, Chang Liu, Xiaosong Zhang, Jianbin Jiao, Xi- angyang Ji, and Qixiang Ye. Beyond bounding-box: Convex- hull feature adaptation for oriented and densely packed ob- ject detection. InProceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition, pages 8792– 8801, 2021. 2

    work page 2021

  7. [7]

    Align deep features for oriented object detection.IEEE transac- tions on geoscience and remote sensing, 60:1–11, 2021

    Jiaming Han, Jian Ding, Jie Li, and Gui-Song Xia. Align deep features for oriented object detection.IEEE transac- tions on geoscience and remote sensing, 60:1–11, 2021. 3, 7

    work page 2021

  8. [8]

    Misaligned rgb-infrared object detection via adaptive dual-discrepancy calibration.Remote Sensing, 15(19):4887, 2023

    Mingzhou He, Qingbo Wu, King Ngi Ngan, Feng Jiang, Fan- man Meng, and Linfeng Xu. Misaligned rgb-infrared object detection via adaptive dual-discrepancy calibration.Remote Sensing, 15(19):4887, 2023. 1

    work page 2023

Show all 44 references
  1. [9]

    Multispectral pedestrian detection: Benchmark dataset and baseline

    Soonmin Hwang, Jaesik Park, Namil Kim, Yukyung Choi, and In So Kweon. Multispectral pedestrian detection: Benchmark dataset and baseline. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 1037–1045, 2015. 1

    2015

  2. [10]

    Llvip: A visible-infrared paired dataset for low- light vision

    Xinyu Jia, Chuang Zhu, Minzhen Li, Wenqi Tang, and Wenli Zhou. Llvip: A visible-infrared paired dataset for low- light vision. In2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), pages 3489–3497,

  3. [11]

    R2cnn: Rota- tional region cnn for orientation robust scene text detection

    Yingying Jiang, Xiangyu Zhu, Xiaobing Wang, Shuli Yang, Wei Li, Hua Wang, Pei Fu, and Zhenbo Luo. R2cnn: Rota- tional region cnn for orientation robust scene text detection. arXiv preprint arXiv:1706.09579, 2017. 2

    2017 arXiv

  4. [12]

    Multispectral pedestrian detection via simultaneous detec- tion and segmentation.arXiv preprint arXiv:1808.04818,

    Chengyang Li, Dan Song, Ruofeng Tong, and Min Tang. Multispectral pedestrian detection via simultaneous detec- tion and segmentation.arXiv preprint arXiv:1808.04818,

    arXiv

  5. [13]

    Illumination-aware faster R-CNN for robust multispectral pedestrian detection.Pattern Recognition, 85:161–171,

    Chenglong Li, Dan Song, Ruofeng Tong, and Ming Tang. Illumination-aware faster R-CNN for robust multispectral pedestrian detection.Pattern Recognition, 85:161–171,

  6. [14]

    Multiscale cross-modal homo- geneity enhancement and confidence-aware fusion for mul- tispectral pedestrian detection.IEEE Transactions on Multi- media, 26:852–863, 2023

    Ruimin Li, Jiajun Xiang, Feixiang Sun, Ye Yuan, Longwu Yuan, and Shuiping Gou. Multiscale cross-modal homo- geneity enhancement and confidence-aware fusion for mul- tispectral pedestrian detection.IEEE Transactions on Multi- media, 26:852–863, 2023. 3

    2023

  7. [15]

    Ori- ented reppoints for aerial object detection

    Wentong Li, Yijie Chen, Kaixuan Hu, and Jianke Zhu. Ori- ented reppoints for aerial object detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1829–1838, 2022. 2

    2022

  8. [16]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll´ar. Focal loss for dense object detection. InPro- ceedings of the IEEE international conference on computer vision, pages 2980–2988, 2017. 7

    2017

  9. [17]

    Como: Cross-mamba interaction and offset- guided fusion for multimodal object detection.Information Fusion, 125:103414, 2026

    Chang Liu, Xin Ma, Xiaochen Yang, Yuxiang Zhang, and Yanni Dong. Como: Cross-mamba interaction and offset- guided fusion for multimodal object detection.Information Fusion, 125:103414, 2026. 7

    2026

  10. [18]

    Multispectral deep neural networks for pedestrian detection

    Jingjing Liu, Shaoting Zhang, Shu Wang, and Dimitris N Metaxas. Multispectral deep neural networks for pedestrian detection. 2016. 1

    2016

  11. [19]

    Arbitrary-oriented scene text detection via rotation proposals.IEEE transac- tions on multimedia, 20(11):3111–3122, 2018

    Jianqi Ma, Weiyuan Shao, Hao Ye, Li Wang, Hong Wang, Yingbin Zheng, and Xiangyang Xue. Arbitrary-oriented scene text detection via rotation proposals.IEEE transac- tions on multimedia, 20(11):3111–3122, 2018. 2

    2018

  12. [20]

    Cross- modality fusion transformer for multispectral object detec- tion.arXiv preprint arXiv:2111.00273, 2021

    Fang Qingyun, Han Dapeng, and Wang Zhaokui. Cross- modality fusion transformer for multispectral object detec- tion.arXiv preprint arXiv:2111.00273, 2021. 3

    2021 arXiv

  13. [21]

    Cross-modality attentive feature fusion for object detection in multispectral remote sensing imagery.Pattern Recognition, 130:108786, 2022

    Fang Qingyun and Wang Zhaokui. Cross-modality attentive feature fusion for object detection in multispectral remote sensing imagery.Pattern Recognition, 130:108786, 2022. 3

    2022

  14. [22]

    Icafusion: Iterative cross-attention guided feature fusion for multispectral object detection.Pattern Recognition, 145:109913, 2024

    Jifeng Shen, Yifei Chen, Yue Liu, Xin Zuo, Heng Fan, and Wankou Yang. Icafusion: Iterative cross-attention guided feature fusion for multispectral object detection.Pattern Recognition, 145:109913, 2024. 3

    2024

  15. [23]

    Misaligned visible- thermal object detection: A drone-based benchmark and baseline.IEEE Transactions on Intelligent Vehicles, 2024

    Kechen Song, Xiaotong Xue, Hongwei Wen, Yingying Ji, Yunhui Yan, and Qinggang Meng. Misaligned visible- thermal object detection: A drone-based benchmark and baseline.IEEE Transactions on Intelligent Vehicles, 2024. 1

    2024

  16. [24]

    Drone-based rgb-infrared cross-modality vehicle detection via uncertainty-aware learning.IEEE Transactions on Cir- cuits and Systems for Video Technology, 32(10):6700–6713,

    Yiming Sun, Bing Cao, Pengfei Zhu, and Qinghua Hu. Drone-based rgb-infrared cross-modality vehicle detection via uncertainty-aware learning.IEEE Transactions on Cir- cuits and Systems for Video Technology, 32(10):6700–6713,

  17. [25]

    Multispectral pedestrian detection using deep fusion convolutional neural networks

    J ¨org Wagner, V olker Fischer, Michael Herman, Sven Behnke, et al. Multispectral pedestrian detection using deep fusion convolutional neural networks. InESANN, volume 587, pages 509–514, 2016. 3

    2016

  18. [26]

    Learning a dynamic cross-modal network for multispectral pedestrian detection

    Jin Xie, Rao Muhammad Anwer, Hisham Cholakkal, Jing Nie, Jiale Cao, Jorma Laaksonen, and Fahad Shahbaz Khan. Learning a dynamic cross-modal network for multispectral pedestrian detection. InProceedings of the 30th ACM In- ternational Conference on Multimedia, pages 4043–4052,

  19. [27]

    Oriented r-cnn for object detection

    Xingxing Xie, Gong Cheng, Jiabao Wang, Xiwen Yao, and Junwei Han. Oriented r-cnn for object detection. InProceed- ings of the IEEE/CVF international conference on computer vision, pages 3520–3529, 2021. 2

    2021

  20. [28]

    Gliding vertex on the horizontal bounding box for multi-oriented object detec- tion.IEEE transactions on pattern analysis and machine intelligence, 43(4):1452–1459, 2020

    Yongchao Xu, Mingtao Fu, Qimeng Wang, Yukang Wang, Kai Chen, Gui-Song Xia, and Xiang Bai. Gliding vertex on the horizontal bounding box for multi-oriented object detec- tion.IEEE transactions on pattern analysis and machine intelligence, 43(4):1452–1459, 2020. 2

    2020

  21. [29]

    Dense label encoding for boundary discontinuity free rotation detection

    Xue Yang, Liping Hou, Yue Zhou, Wentao Wang, and Junchi Yan. Dense label encoding for boundary discontinuity free rotation detection. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 15819–15829, 2021. 3

    2021

  22. [30]

    Arbitrary-oriented object detec- tion with circular smooth label

    Xue Yang and Junchi Yan. Arbitrary-oriented object detec- tion with circular smooth label. InEuropean conference on computer vision, pages 677–694. Springer, 2020. 3

    2020

  23. [31]

    R3det: Refined single-stage detector with feature refinement for ro- tating object

    Xue Yang, Junchi Yan, Ziming Feng, and Tao He. R3det: Refined single-stage detector with feature refinement for ro- tating object. InProceedings of the AAAI conference on ar- tificial intelligence, volume 35, pages 3163–3171, 2021. 2, 7

    2021

  24. [32]

    Rethinking rotated object detection with gaussian wasserstein distance loss

    Xue Yang, Junchi Yan, Qi Ming, Wentao Wang, Xiaopeng Zhang, and Qi Tian. Rethinking rotated object detection with gaussian wasserstein distance loss. InInternational con- ference on machine learning, pages 11830–11841. PMLR,

  25. [33]

    Learning high-precision bounding box for rotated object detection via kullback- leibler divergence.Advances in Neural Information Process- ing Systems, 34:18381–18394, 2021

    Xue Yang, Xiaojiang Yang, Jirui Yang, Qi Ming, Wentao Wang, Qi Tian, and Junchi Yan. Learning high-precision bounding box for rotated object detection via kullback- leibler divergence.Advances in Neural Information Process- ing Systems, 34:18381–18394, 2021. 3

    2021

  26. [34]

    The kfiou loss for rotated object detection.arXiv preprint arXiv:2201.12558, 2022

    Xue Yang, Yue Zhou, Gefan Zhang, Jirui Yang, Wentao Wang, Junchi Yan, Xiaopeng Zhang, and Qi Tian. The kfiou loss for rotated object detection.arXiv preprint arXiv:2201.12558, 2022. 3, 7

    2022 arXiv

  27. [35]

    Improving rgb-infrared object detection with cascade alignment-guided transformer.Information Fusion, 105:102246, 2024

    Maoxun Yuan, Xiaorong Shi, Nan Wang, Yinyan Wang, and Xingxing Wei. Improving rgb-infrared object detection with cascade alignment-guided transformer.Information Fusion, 105:102246, 2024. 1

    2024

  28. [36]

    Transla- tion, scale and rotation: cross-modal alignment meets rgb- infrared vehicle detection

    Maoxun Yuan, Yinyan Wang, and Xingxing Wei. Transla- tion, scale and rotation: cross-modal alignment meets rgb- infrared vehicle detection. InEuropean Conference on Com- puter Vision, pages 509–525. Springer, 2022. 1

    2022

  29. [37]

    C 2former: Calibrated and complementary transformer for rgb-infrared object de- tection.IEEE Transactions on Geoscience and Remote Sens- ing, 62:1–12, 2024

    Maoxun Yuan and Xingxing Wei. C 2former: Calibrated and complementary transformer for rgb-infrared object de- tection.IEEE Transactions on Geoscience and Remote Sens- ing, 62:1–12, 2024. 7

    2024

  30. [38]

    Weakly aligned feature fusion for multimodal object detection.IEEE Transactions on Neural Networks and Learning Systems, 2021

    Lu Zhang, Zhiyong Liu, Xiangyu Zhu, Zhan Song, Xu Yang, Zhen Lei, and Hong Qiao. Weakly aligned feature fusion for multimodal object detection.IEEE Transactions on Neural Networks and Learning Systems, 2021. 1

    2021

  31. [39]

    Weakly aligned cross-modal learn- ing for multispectral pedestrian detection

    Lu Zhang, Xiangyu Zhu, Xiangyu Chen, Xu Yang, Zhen Lei, and Zhiyong Liu. Weakly aligned cross-modal learn- ing for multispectral pedestrian detection. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2019. 1

    2019

  32. [40]

    Tfdet: Target-aware fusion for rgb-t pedestrian detection

    Xue Zhang, Xiaohan Zhang, Jiangtao Wang, Jiacheng Ying, Zehua Sheng, Heng Yu, Chunguang Li, and Hui-Liang Shen. Tfdet: Target-aware fusion for rgb-t pedestrian detection. IEEE Transactions on Neural Networks and Learning Sys- tems, 36(7):13276–13290, 2024. 3

    2024

  33. [41]

    Removal then selection: A coarse-to-fine fusion perspective for rgb-infrared object detection.IEEE Transactions on Intelligent Transportation Systems, 2025

    Tianyi Zhao, Maoxun Yuan, Feng Jiang, Nan Wang, and Xingxing Wei. Removal then selection: A coarse-to-fine fusion perspective for rgb-infrared object detection.IEEE Transactions on Intelligent Transportation Systems, 2025. 3

    2025

  34. [42]

    Gfd-ssd: Gated fusion double ssd for multispectral pedestrian detec- tion.arXiv preprint arXiv:1903.06999, 2019

    Yang Zheng, Izzat H Izzat, and Shahrzad Ziaee. Gfd-ssd: Gated fusion double ssd for multispectral pedestrian detec- tion.arXiv preprint arXiv:1903.06999, 2019. 3

    1903 arXiv

  35. [43]

    Improving multi- spectral pedestrian detection by addressing modality imbal- ance problems

    Kailai Zhou, Linsen Chen, and Xun Cao. Improving multi- spectral pedestrian detection by addressing modality imbal- ance problems. InEuropean conference on computer vision, pages 787–803. Springer, 2020. 3

    2020

  36. [44]

    Dmm: Disparity-guided multispectral mamba for oriented object detection in remote sensing.IEEE Transactions on Geoscience and Remote Sensing, 63:1–13,

    Minghang Zhou, Tianyu Li, Chaofan Qiao, Dongyu Xie, Guoqing Wang, Ningjuan Ruan, Lin Mei, Yang Yang, and Heng Tao Shen. Dmm: Disparity-guided multispectral mamba for oriented object detection in remote sensing.IEEE Transactions on Geoscience and Remote Sensing, 63:1–13,

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.