Pith. sign in

REVIEW 3 major objections 6 minor 73 references

InfraNet trains infrared detectors with quality-gated RGB help, then drops RGB for free IR-only inference.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-11 23:53 UTC pith:EKEEHX5H

load-bearing objection Clean IR-centric training/export pattern with solid multi-benchmark numbers; the scalar QualGate is simple but the ablations hold. the 3 major comments →

arxiv 2607.03795 v1 pith:EKEEHX5H submitted 2026-07-04 cs.CV

InfraNet: Quality-Aware RGB Guidance for Efficient Infrared Object Detection

Zichao Feng , Haodong Zhu , Jingying Yang , Sheng Xu , Yangyang Ren , Yuguang Yang , Xuhui Liu , Juan Zhang
show 3 more authors
Tian Wang Linlin Yang Baochang Zhang
This is my paper
classification cs.CV
keywords infrared object detectionquality-aware fusionmodality-asymmetric learningRGB-IRprivileged informationQualGateIR-only deployment
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Standard RGB–infrared detectors fuse both streams equally and still need both cameras at test time, so a dark or foggy RGB image can hurt the very system it is meant to help. InfraNet reverses the priority: the infrared branch is the primary path that produces the final detections, while RGB is treated only as optional training-time guidance. A lightweight gate called QualGate predicts a single reliability score from the RGB features; when that score is low it down-weights the RGB contribution and simultaneously boosts the infrared features so the network never learns from contaminated cues. After training, the entire RGB branch and all gates can be stripped away, leaving a pure infrared detector whose accuracy still reflects the privileged RGB supervision. On four standard benchmarks the resulting IR-only models match or beat heavier dual-modal systems while running at the same cost as ordinary single-modal detectors.

Core claim

An asymmetric, IR-centric training scheme that uses RGB solely as reliability-controlled auxiliary supervision can produce infrared detectors that are both more accurate under adverse lighting and free of any dual-modal cost at inference. The same architecture can keep the RGB stream when it is available, giving a higher-capacity dual-modal detector under the same quality-aware regulation.

What carries the argument

QualGate: a multi-scale fusion module that predicts a scalar reliability score q from global-average-pooled RGB features, multiplies the RGB stream by q (suppression) and multiplies the IR stream by a clipped inverse gain α(q) (compensation), thereby shaping IR-centered representations without external quality labels.

Load-bearing premise

A single scalar score computed from a global pool of RGB features, together with a fixed hand-chosen amplification schedule, is enough to decide when RGB is helpful and when it should be ignored.

What would settle it

Train identical InfraNet models with and without QualGate on a night-only split where RGB is heavily degraded; if the gated IR-only model no longer outperforms a pure-IR baseline or a naïve equal-weight fusion baseline on that split, the quality-control claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • IR-only detectors can inherit the accuracy gains of multi-modal training without carrying multi-modal compute or sensor cost at deployment.
  • When both cameras remain available, the same quality gate can be left active, yielding a higher-capacity dual-modal detector under one training recipe.
  • Negative transfer from degraded RGB can be treated as a controllable training signal rather than an inevitable fusion risk.
  • Any multi-scale detector family can host the same asymmetric QualGate pattern, so the idea is not tied to a single backbone.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same scalar-gate idea could be applied to other privileged-modality pairs (depth, event cameras, radar) where one stream is cheap at inference and the other is only intermittently reliable.
  • Because the gate never requires explicit quality labels, the method may transfer to unlabeled or weakly paired multi-modal corpora that lack night/day annotations.
  • If the learned score q correlates with scene illumination, it could later serve as a cheap online sensor-health signal for adaptive fusion policies.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes InfraNet, an IR-centric multi-modal detection framework that treats RGB as reliability-controlled auxiliary guidance rather than a co-equal modality. Its core module, QualGate (§3.1, Eqs. 1–4), predicts a scalar task-oriented reliability score q from GAP of the auxiliary features and uses it both to suppress RGB contributions and to amplify multi-scale IR features via a clipped inverse gain α(q). Two independently trained variants are defined: InfraNet-IR (RGB used only at training; pure IR backbone–neck–head at inference) and InfraNet-RGB-IR (dual-stream at both train and test). Experiments on LLVIP, FLIR-Aligned, M3FD and DroneVehicle across ResNet50 and YOLO-family backbones report competitive or superior mAP/mAP50 relative to recent fusion methods (WaveMamba, CFT, etc.), with IR-only models retaining the parameter/latency footprint of a single-stream detector. Ablations isolate RGB suppression vs. IR amplification (Table 5), controlled RGB degradation behavior of q (Table 6), and the auxiliary loss weight w_aux (Table 7).

Significance. If the reported numbers hold under independent re-implementation, the work supplies a practical privileged-guidance recipe that decouples training-time multi-modal supervision from test-time modality availability—an important deployment axis for thermal perception under night, fog and UAV conditions. Strengths that raise the contribution above a pure engineering tweak include: (i) consistent gains across four public RGB–IR benchmarks and multiple detector families, (ii) explicit zero-overhead IR-only export, (iii) controlled degradation sweeps that show q responds to RGB corruption without external quality labels, and (iv) component ablations that separate suppression from amplification. The design is simple enough to be portable; the main novelty is the asymmetric training/inference contract rather than a new backbone.

major comments (3)
  1. §5.2 / Tables 1–4 report only single-run point estimates with no standard deviations, multiple seeds, or statistical tests. Given that several claimed margins over WaveMamba are modest (e.g., LLVIP YOLOv8 mAP 70.5 vs 66.0; FLIR mAP gaps of a few points), the ranking of “strong or competitive” cannot be assessed for stability. At minimum the authors should re-run the primary YOLO/ResNet configurations with 3 seeds and report mean±std for mAP50 and mAP, or supply a clear statement that only single-run numbers are available and why.
  2. §3.1 Eqs. (2)–(4) and the accompanying text leave the IR amplification schedule under-specified for reproducibility. α(q)=clip(1.5−q;1.0,1.5) is acknowledged as non-unique, yet the paper does not report a sensitivity sweep over the upper bound (or alternative functional forms) in the main text; the supplementary is only alluded to. Because the dual-path design is presented as load-bearing (Table 5), a short main-text ablation of the clip range (or a learned α) is needed so that readers can judge how tightly the reported gains depend on the hand-chosen constant.
  3. §5.2 Table 4 (M3FD Day/Overcast/Night/Challenge) is labeled a “diagnostic comparison” and notes that many recent methods do not report the same subset protocol, yet the table still ranks InfraNet first by large margins against older fusion baselines. Either the authors should re-evaluate the strongest recent competitors (WaveMamba, RSDet, etc.) under the identical subset split, or they should move Table 4 to the appendix and avoid presenting it as a primary SOTA comparison. As written, the subset numbers risk overstating the method’s advantage.
minor comments (6)
  1. Fig. 1 caption and surrounding text claim that CFT/FusionMamba/WaveMamba “introduce additional noise” relative to pure IR; the visualization is qualitative channel-sum heatmaps. A quantitative noise or SNR proxy (or at least a clearer description of how the maps were normalized) would make the motivation more rigorous.
  2. §3.2 / Fig. 2: the description of which features ({A3,A4,A5} vs {P3,P4,P5}) feed the detection head during training for InfraNet-IR is slightly ambiguous; a one-sentence clarification of the dual-head loss paths would help re-implementation.
  3. Notation: Ain is used both for “RGB feature or fused feature” (§3.1) and later as the auxiliary input at each site; a short glossary or consistent subscripting would reduce confusion.
  4. Related work (§2.2) cites HalluciDet, CRLM and CAGF; Table 5 already replaces QualGate with those modules, which is good, but a one-line discussion of why a scalar q is preferred over their denser reliability maps would strengthen the positioning.
  5. Minor typos / style: “M3FD” vs “M$^3$FD” inconsistency; “compensateIRfeaturesduringcross-modaltraining” spacing artifacts in the abstract; “tip aggregation” in the DroneVehicle discussion appears to be a typo for “feature aggregation”.
  6. Inference-time numbers in Table 2 (ms) are useful; stating the batch size and whether TensorRT/ONNX was used would improve comparability.

Circularity Check

0 steps flagged

No significant circularity: empirical detector design whose claims rest on external benchmarks and ablations, not on self-definitional or fitted-as-prediction reductions.

full rationale

InfraNet is a standard empirical multi-modal detection paper. QualGate (Eqs. 1–4) defines a learned scalar q from GAP(A_in) and a hand-chosen IR gain α(q)=clip(1.5−q;1.0,1.5); the paper itself states that 1.5 is not uniquely optimal. The central claims are measured mAP/mAP50 on LLVIP, FLIR-Aligned, M3FD and DroneVehicle for both IR-only and RGB–IR deployments (Tables 1–4), plus ablations that remove suppression/amplification, replace QualGate with CAGF/CRLM/HalluciDet, vary w_aux, and sweep controlled RGB fog/brightness (Tables 5–7). None of these results is forced by construction from a fitted parameter renamed as a prediction, nor does any uniqueness theorem or load-bearing self-citation close the argument. Related-work citations (including WaveMamba and HalluciDet) are used for comparison baselines, not as premises that make the reported numbers true by definition. The derivation chain is therefore self-contained against external benchmarks; score 0 is appropriate.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 2 invented entities

Claims rest on the invented QualGate control mechanism, a handful of hand-chosen scalars selected by ablation, and the domain premise that IR should be the primary modality. No external quality annotations or theoretical guarantees are supplied for q.

free parameters (3)
  • w_aux = 0.25
    Auxiliary detection-loss weight; default 0.25 chosen by ablation on LLVIP and used for all main results (Table 7).
  • alpha_upper = 1.5
    Upper clip value in the IR amplification rule α(q)=clip(1.5−q;1.0,1.5); paper states it is not uniquely optimal (§3.1).
  • QualGate MLP depth/width = two-layer bottleneck
    Two-layer bottleneck networks for q and content gates; architectural free choice for efficiency.
axioms (3)
  • domain assumption Under adverse illumination or weather the infrared modality is more reliable than RGB and should therefore be the primary pathway.
    Stated in the introduction and used to justify the entire asymmetric design.
  • ad hoc to paper A single scalar q obtained from global-average-pooled RGB features via a small MLP is an adequate task-oriented reliability control signal.
    Core modeling choice of QualGate (Eq. 1); no external quality labels or richer estimators are used.
  • domain assumption Standard YOLO detection losses, assignment rules and 300-epoch SGD schedules transfer unchanged to the multi-modal setting.
    Adopted without modification for all experiments (§4–5).
invented entities (2)
  • QualGate no independent evidence
    purpose: Learns scalar q to suppress unreliable RGB guidance and amplify IR features at multiple scales.
    New module introduced in §3.1; validated only by ablations inside this paper.
  • InfraNet / InfraNet-IR / InfraNet-RGB-IR no independent evidence
    purpose: Asymmetric IR-primary architectures that support either IR-only or dual-modal deployment after quality-aware training.
    Proposed framework whose performance numbers constitute the central empirical claim.

pith-pipeline@v1.1.0-grok45 · 21276 in / 2590 out tokens · 36140 ms · 2026-07-11T23:53:55.337906+00:00 · methodology

0 comments
read the original abstract

Robust object detection under adverse visual conditions remains a long-standing challenge for multi-modal perception systems. Existing fusion-based methods typically require both RGB and infrared (IR) inputs, and treat them equally during both training and inference, which compromises their robustness when the RGB modality becomes unreliable or unavailable. In this case, we propose \textbf{InfraNet}, an IR-centric quality-aware framework that regulates RGB guidance during training and supports flexible RGB--IR or IR-only deployment. InfraNet employs an asymmetric architecture where the primary IR pathway extracts multi-scale infrared features for predictions, while the auxiliary RGB pathway provides reliability-controlled supervisory signals. The core of InfraNet is \textbf{QualGate}, a quality-aware fusion module that learns a task-oriented control signal to suppress unreliable RGB guidance and compensate IR features during cross-modal training. Built upon InfraNet, we design two architectural variants: a lightweight IR-only architecture InfraNet-IR and an RGB--IR architecture InfraNet-RGB-IR. Our method is evaluated through extensive experiments on four benchmark datasets (LLVIP, FLIR-Aligned, M$^3$FD, and DroneVehicle), showing strong or competitive accuracy in challenging low-light and adverse weather conditions. Notably, InfraNet maintains high efficiency in IR-only inference, making it both accurate and computationally efficient.

Figures

Figures reproduced from arXiv: 2607.03795 by Baochang Zhang, Haodong Zhu, Jingying Yang, Juan Zhang, Linlin Yang, Sheng Xu, Tian Wang, Xuhui Liu, Yangyang Ren, Yuguang Yang, Zichao Feng.

Figure 1
Figure 1. Figure 1: Visualization of RGB and IR feature quality under nighttime condi￾tions on M3FD dataset. While RGB fea￾tures exhibit substantial noise and degra￾dation, IR features remain robust and clear. Existing fusion methods (CFT, Fu￾sionMamba, WaveMamba) introduce ad￾ditional noise compared to pure IR fea￾tures, motivating our IR-centric design that regulates RGB guidance during train￾ing. While significant progress… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of InfraNet-IR (YOLOv9-based). During training, a main IR branch (black) and an auxiliary RGB branch (blue) interact via QualGate modules at three fusion sites (A3, A4, A5). Through quality-aware suppression/amplification, QualGate constrains the cross-branch interaction to promote IR-consistent representations while preventing degraded RGB cues from contaminating the IR stream, which justifies re… view at source ↗
Figure 3
Figure 3. Figure 3: Architecture of QualGate. It takes as input K scales IR features {Ck} K k=1 and the auxiliary feature Ain, and targets a quality-aware fused output Aout. For fusion, the control branch predicts a scalar score q for auxiliary guidance and compensates IR features accordingly based on α(q). 3.2 InfraNet To avoid degraded RGB contaminating informative IR features and inducing negative transfer, we propose Infr… view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of detection results of several cross-modality object detection methods on M3FD dataset. InfraNet-IR produces tighter bounding boxes, fewer false positives, and more consistent detections across diverse illumination conditions. Unlike prior methods (SuperFusion, TarDAL, CFT, FusionMamba), which often suffer from RGB degradation and background leakage, InfraNet-IR maintains robust localization… view at source ↗
Figure 5
Figure 5. Figure 5: Feature-level visualization on M3FD nighttime examples. InfraNet-IR produces more target-aligned responses with less background activation. This visualization is used to diagnose the effect of RGB-guided training under degraded RGB conditions; the deployed InfraNet-IR model still performs IR-only inference. spheric scattering at high altitudes—a critical advantage for aerial imaging where RGB visibility is… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

73 extracted references · 6 linked inside Pith

  1. [1]

    Sensors (2023)

    Bao, C., Cao, J., Hao, Q., Cheng, Y., Ning, Y., Zhao, T.: Dual-yolo architecture from infrared and visible images for object detection. Sensors (2023)

    2023

  2. [2]

    In: CVPR

    Bijelic, M., Gruber, T., Mannan, F., Kraus, F., Ritter, W., Dietmayer, K., Heide, F.: Seeing through fog without seeing fog: Deep multimodal sensor fusion in unseen adverse weather. In: CVPR. pp. 11682–11692 (2020)

    2020

  3. [3]

    In: CVPRW (2023)

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

    2023

  4. [4]

    In: CVPR (2024)

    Chen, C., Qi, J., Liu, X., Bin, K., Fu, R., Hu, X., Zhong, P.: Weakly misalignment- free adaptive feature alignment for uavs-based multimodal object detection. In: CVPR (2024)

    2024

  5. [5]

    In: ECCV (2022)

    Chen, Y.T., Shi, J., Ye, Z., Mertz, C., Ramanan, D., Kong, S.: Multimodal object detection via probabilistic ensembling. In: ECCV (2022)

    2022

  6. [6]

    Multimedia Tools Appl

    Cheng, X., Geng, K., Wang, Z., Wang, J., Sun, Y., Ding, P.: Slbaf-net: Super- lightweight bimodal adaptive fusion network for uav detection in low recognition environment. Multimedia Tools Appl. (2023)

    2023

  7. [7]

    In: WACV

    Deevi, S.A., Lee, C., Gan, L., Nagesh, S., Pandey, G., Chung, S.J.: Rgb-x object detection via scene-specific fusion modules. In: WACV. pp. 7366–7375 (2024)

    2024

  8. [8]

    IEEE Trans

    Dong, W., Zhu, H., Lin, S., Luo, X., Shen, Y., Guo, G., Zhang, B.: Fusion-mamba for cross-modality object detection. IEEE Trans. Multim.27, 7392–7406 (2025)

    2025

  9. [9]

    arXiv preprint arXiv:2111.00273 (2021)

    Fang, Q., Han, D., Wang, Z.: Cross-modality fusion transformer for multispectral object detection. arXiv preprint arXiv:2111.00273 (2021)

    Pith/arXiv arXiv 2021

  10. [10]

    Fei, X., Guo, M., Li, Y., Yu, R., Sun, L.: Acdf-yolo: Attentive and cross-differential fusion network for multimodal remote sensing object detection. Remote. Sens. (2024)

    2024

  11. [11]

    Gui, S., Song, S., Qin, R., Tang, Y.: Remote sensing object detection in the deep learning era - A review. Remote. Sens.16(2), 327 (2024)

    2024

  12. [12]

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition (2015),https://arxiv.org/abs/1512.03385, accessed: June 29, 2026

    Pith/arXiv arXiv 2015

  13. [13]

    arXiv preprint arXiv:1503.02531 (2015)

    Hinton, G., Vinyals, O., Dean, J.: Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531 (2015)

    Pith/arXiv arXiv 2015

  14. [14]

    IEEE Access12, 42816–42833 (2024)

    Hussain,M.:Yolov1tov8:Unveilingeachvariant-acomprehensivereviewofYOLO. IEEE Access12, 42816–42833 (2024)

    2024

  15. [15]

    In: ICCVW (2021)

    Jia, X., Zhu, C., Li, M., Tang, W., Zhou, W.: Llvip: A visible-infrared paired dataset for low-light vision. In: ICCVW (2021)

    2021

  16. [16]

    Jiang, C., Ren, H., Yang, H., Huo, H., Zhu, P., Yao, Z., Li, J., Sun, M., Yang, S.: M2fnet: Multi-modal fusion network for object detection from visible and thermal infrared images. Int. J. Appl. Earth Obs. Geoinf. (2024)

    2024

  17. [17]

    2-yolov5 classification models, apple ml, reproducibility, clearml and deci

    Jocher, G., chaurasia, A., Stoken, A., Borovec, J., Kwon, Y., Michael, K., Fang, J., Wong, c., Yifu, Z., Montes, D., et al.: ultralytics/yolov5: v6. 2-yolov5 classification models, apple ml, reproducibility, clearml and deci. ai integrations. Zenodo (2022) 16 Feng et al

    2022

  18. [18]

    IEEE Geosci

    Kang, X., Yin, H., Duan, P.: Global–local feature fusion network for visi- ble–infrared vehicle detection. IEEE Geosci. Remote Sens. Lett. (2024)

    2024

  19. [19]

    IEEE Access8, 125459–125476 (2020)

    Kristo, M., Ivasic-Kos, M., Pobar, M.: Thermal object detection in difficult weather conditions using YOLO. IEEE Access8, 125459–125476 (2020)

    2020

  20. [20]

    Pattern Recognit

    Lee, S., Park, J., Park, J.: Crossformer: Cross-guided attention for multi-modal object detection. Pattern Recognit. Lett. (2024)

    2024

  21. [21]

    IEEE Trans

    Li, A., Wang, Z., Wang, F., Liu, Z., Yin, G., Fang, R., Geng, K.: A novel semantic information perception architecture for extreme targets detection in complex traffic scenarios. IEEE Trans. Intell. Veh. (2024)

    2024

  22. [22]

    PR (2019)

    Li, C., Song, D., Tong, R., Tang, M.: Illumination-aware faster r-cnn for robust multispectral pedestrian detection. PR (2019)

    2019

  23. [23]

    Expert Systems with Applications p

    Li, H., et al.: Crossmodalnet: A dual-modal object detection network based on cross-modal fusion and channel interaction. Expert Systems with Applications p. 129677 (2025)

    2025

  24. [24]

    IEEE Transactions on Image Processing28(5), 2614–2623 (2019)

    Li, H., Wu, X.J., Kittler, J.: Densefuse: A fusion approach to infrared and visible images. IEEE Transactions on Image Processing28(5), 2614–2623 (2019)

    2019

  25. [25]

    IEEE Access7, 103964–103977 (2019)

    Li, H., Wu, X.J., Kittler, J.: Infrared and visible image fusion using a dual discrim- inator conditional generative adversarial network. IEEE Access7, 103964–103977 (2019)

    2019

  26. [26]

    In: ACMMM (2023)

    Li, J., Chen, J., Liu, J., Ma, H.: Learning a graph neural network with cross modality interaction for image fusion. In: ACMMM (2023)

    2023

  27. [27]

    TCSVT (2024)

    Li, Q., Zhang, C., Hu, Q., Zhu, P., Fu, H., Chen, L.: Stabilizing multispectral pedestrian detection with evidential hybrid fusion. TCSVT (2024)

    2024

  28. [28]

    IEEE Trans

    Li, X., Chen, S., Tian, C., Zhou, H., Zhang, Z.: M2fnet: Mask-guided multi-level fusion for rgb-t pedestrian detection. IEEE Trans. Multimed. (2024)

    2024

  29. [29]

    In: CVPR (2022)

    Liu, J., Fan, X., Huang, Z., Wu, G., Liu, R., Zhong, W., Luo, Z.: Target-aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection. In: CVPR (2022)

    2022

  30. [30]

    IEEE Transactions on Circuits and Systems for Video Technology32(1), 105–119 (2022)

    Liu, J., Fan, X., Jiang, J., Liu, R., Luo, Z.: Learning a deep multi-scale feature ensemble and an edge-attention guidance for image fusion. IEEE Transactions on Circuits and Systems for Video Technology32(1), 105–119 (2022)

    2022

  31. [31]

    In: ACMMM

    Liu, L., Li, C., Xiao, Y., Tang, J.: Quality-aware rgbt tracking via supervised reliability learning and weighted residual guidance. In: ACMMM. pp. 3129–3137 (2023)

    2023

  32. [32]

    Information Fusion48, 11–26 (2019)

    Ma, J., Yu, W., Liang, P., Jiang, J., Ling, H.: Fusiongan: A generative adversarial network for infrared and visible image fusion. Information Fusion48, 11–26 (2019)

    2019

  33. [33]

    IEEE Transactions on Instru- mentation and Measurement70, 1–14 (2020)

    Ma, J., Zhang, H., Shao, Z., Liang, P., Xu, H.: Ganmcc: A generative adversarial network with multiclassification constraints for ivif. IEEE Transactions on Instru- mentation and Measurement70, 1–14 (2020)

    2020

  34. [34]

    In: WACV (2025)

    Medeiros, H.R., et al.: Mixed patch visible-infrared modality agnostic object de- tection. In: WACV (2025)

    2025

  35. [35]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

    Medeiros, H.R., Peña, F.A.G., Aminbeidokhti, M., Dubail, T., Granger, E., Ped- ersoli, M.: Hallucidet: Hallucinating rgb modality for person detection through privileged information. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). pp. 1444–1453 (2024)

    2024

  36. [36]

    Payghode, V., Goyal, A., Bhan, A., Iyer, S.S., Dubey, A.K.: Object detection and activity recognition in video surveillance using neural networks. Int. J. Web Inf. Syst.19(3/4), 123–138 (2023)

    2023

  37. [37]

    PR (2024) InfraNet for Infrared Object Detection 17

    Shen, J., Chen, Y., Liu, Y., Zuo, X., Fan, H., Yang, W.: Icafusion: Iterative cross- attention guided feature fusion for multispectral object detection. PR (2024) InfraNet for Infrared Object Detection 17

    2024

  38. [38]

    Remote Sens

    Sun, X., Yu, Y., Cheng, Q.: Adaptive multimodal feature fusion with frequency domain gate for remote sensing object detection. Remote Sens. Lett. (2024)

    2024

  39. [39]

    IEEE Trans

    Sun, X., Yu, Y., Cheng, Q.: Low-rank multimodal remote sensing object detection with frequency filtering experts. IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  40. [40]

    In: ACMMM (2022)

    Sun, Y., Cao, B., Zhu, P., Hu, Q.: Detfusion: A detection-driven infrared and visible image fusion network. In: ACMMM (2022)

    2022

  41. [41]

    TCSVT (2022)

    Sun, Y., Cao, B., Zhu, P., Hu, Q.: Drone-based rgb-infrared cross-modality vehicle detection via uncertainty-aware learning. TCSVT (2022)

    2022

  42. [42]

    IEEE/CAA J

    Tang, L., Deng, Y., Ma, Y., Huang, J., Ma, J.: Superfusion: A versatile image registration and fusion network with semantic awareness. IEEE/CAA J. Autom. Sin. (2022)

    2022

  43. [43]

    Tang, L., Xiang, X., Zhang, H., Gong, M., Ma, J.: Divfusion: Darkness-free infrared and visible image fusion. Inf. Fusion (2023)

    2023

  44. [44]

    flir.com/oem/adas/adas-dataset-form/(2018), accessed: June 29, 2026

    Teledyne FLIR: Free flir thermal dataset for algorithm training.https://www. flir.com/oem/adas/adas-dataset-form/(2018), accessed: June 29, 2026

    2018

  45. [45]

    Sensors (2024)

    Tian, D., Yan, X., Zhou, D., Wang, C., Zhang, W.: Iv-yolo: A lightweight dual- branch object detection network. Sensors (2024)

    2024

  46. [46]

    Neural Networks22(5-6), 544–557 (2009)

    Vapnik, V., Vashist, A.: A new learning paradigm: Learning using privileged infor- mation. Neural Networks22(5-6), 544–557 (2009)

    2009

  47. [47]

    In: ADICS (2024)

    Varghese, R., M., S.: Yolov8: A novel object detection algorithm with enhanced performance and robustness. In: ADICS (2024)

    2024

  48. [48]

    arXiv preprint arXiv:2402.13616 (2024)

    Wang, C.Y., Yeh, I.H., Liao, H.Y.M.: Yolov9: Learning what you want to learn using programmable gradient information. arXiv preprint arXiv:2402.13616 (2024)

    Pith/arXiv arXiv 2024

  49. [49]

    Wang, H., Wang, C., Fu, Q., Si, B., Zhang, D., Kou, R., Yu, Y., Feng, C.: Yolofiv: Object detection algorithm for around-the-clock aerial remote sensing images by fusing infrared and visible features. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. (2024)

    2024

  50. [50]

    IEEE Trans

    Wang, H., Wang, C., Fu, Q., Zhang, D., Kou, R., Yu, Y., Song, J.: Cross-modal oriented object detection of uav aerial images based on image feature. IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  51. [51]

    Remote Sens

    Wang, J., Su, N., Zhao, C., Yan, Y., Feng, S.: Multi-modal object detection method based on dual-branch asymmetric attention backbone and feature fusion pyramid network. Remote Sens. (2024)

    2024

  52. [52]

    Wang, J., Xu, C., Zhao, C., Gao, L., Wu, J., Yan, Y., Feng, S., Su, N.: Multimodal object detection of uav remote sensing based on joint representation optimization and specific information enhancement. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. (2024)

    2024

  53. [53]

    In: Proceedings of the 31st ACM International Conference on Multi- media

    Wang, Z., Colonnier, F., Zheng, J., Acharya, J., Jiang, W., Huang, K.: Tirdet: Mono-modality thermal infrared object detection based on prior thermal-to-visible translation. In: Proceedings of the 31st ACM International Conference on Multi- media. pp. 2663–2672 (2023)

    2023

  54. [54]

    IEEE Geoscience and Remote Sensing Letters (2024)

    Xu, F., Xu, T., Hong, L., Peng, P., Guo, J., Li, J.: Enhanced spectral–spatial fusion network for multispectral object detection in ground-aerial images. IEEE Geoscience and Remote Sensing Letters (2024)

    2024

  55. [55]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 44(1), 502–518 (2022)

    Xu, H., Ma, J., Jiang, J., Guo, X., Ling, H.: U2fusion: A unified unsupervised image fusion network. IEEE Transactions on Pattern Analysis and Machine Intelligence 44(1), 502–518 (2022)

    2022

  56. [56]

    In: CVPR (2022)

    Xu, H., Ma, J., Yuan, J., Le, Z., Liu, W.: Rfnet: Unsupervised network for mutually reinforcing multi-modal image registration and fusion. In: CVPR (2022)

    2022

  57. [57]

    In: YAC (2024) 18 Feng et al

    Xu, J., Mo, B., Zhao, J., Zhao, C., Tao, Y., Han, S.: Cross-modal adaptive fusion object detection based on illumination-awareness. In: YAC (2024) 18 Feng et al

    2024

  58. [58]

    arXiv preprint arXiv:2508.07803 (2025)

    Xu, Y., et al.: Mambatrans: Multimodal fusion image translation via large language model priors for downstream visual tasks. arXiv preprint arXiv:2508.07803 (2025)

    Pith/arXiv arXiv 2025

  59. [59]

    World Electric Vehicle Journal16(1), 20 (2025)

    Yang, B., Li, J., Zeng, T.: A review of environmental perception technology based on multi-sensor information fusion in autonomous driving. World Electric Vehicle Journal16(1), 20 (2025)

    2025

  60. [60]

    In: ICHMS (2024)

    Yang, H., Bian, W., Wang, J., Bo, Y., Mi, Y.: A dual-modality pedestrian detection method based on multi-scale feature fusion. In: ICHMS (2024)

    2024

  61. [61]

    In: CVPR (2024)

    Yi, X., Xu, H., Zhang, H., Tang, L., Ma, J.: Text-if: Leveraging semantic text guidance for degradation-aware and interactive image fusion. In: CVPR (2024)

    2024

  62. [62]

    In: ECCV (2022)

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

    2022

  63. [63]

    IEEE Trans

    Yuan, M., Wei, X.: C²former: Calibrated and complementary transformer for rgb- infrared object detection. IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  64. [64]

    IJCV (2021)

    Zhang, H., Ma, J.: Sdnet: A versatile squeeze-and-decomposition network for real- time image fusion. IJCV (2021)

    2021

  65. [65]

    In: CVPR

    Zhang, Q., Zhao, S., Luo, Y., Zhang, D., Huang, N., Han, J.: Abmdrnet: Adaptive- weighted bi-directional modality difference reduction network for rgb-t semantic segmentation. In: CVPR. pp. 2633–2642 (2021)

    2021

  66. [66]

    IEEE Sensors Journal23(24), 30751–30764 (2023)

    Zhao, H., Chu, K., Zhang, J., Feng, C.: Yolo-fsd: An improved target detection algorithm on remote-sensing images. IEEE Sensors Journal23(24), 30751–30764 (2023)

    2023

  67. [67]

    Sensors (2024)

    Zhao, P., Ye, X., Du, Z.: Object detection in multispectral remote sensing images based on cross-modal cross-attention. Sensors (2024)

    2024

  68. [68]

    arXiv preprint arXiv:2401.10731 (2024)

    Zhao, T., Yuan, M., Jiang, F., Wang, N., Wei, X.: Removal then selection: A coarse-to-fine fusion perspective for rgb-infrared object detection. arXiv preprint arXiv:2401.10731 (2024)

    arXiv 2024

  69. [69]

    In: CVPR (2023)

    Zhao, Z., Bai, H., Zhang, J., Zhang, Y., Xu, S., Lin, Z., Timofte, R., Van Gool, L.: Cddfuse: Correlation-driven dual-branch feature decomposition for multi-modality image fusion. In: CVPR (2023)

    2023

  70. [70]

    In: IJCAI (2020)

    Zhao, Z., Xu, S., Zhang, C., Liu, J., Zhang, J., Li, P.: Didfuse: Deep image decom- position for infrared and visible image fusion. In: IJCAI (2020)

    2020

  71. [71]

    arXiv:2407.08132 (2024)

    Zhou, M., Li, T., Qiao, C., Xie, D., Wang, G., Ruan, N., Mei, L., Yang, Y.: Dmm: Disparity-guided multispectral mamba for oriented object detection in remote sens- ing. arXiv:2407.08132 (2024)

    Pith/arXiv arXiv 2024

  72. [72]

    ICCV (2025)

    Zhu, H., Dong, W., Yang, L., Li, H., Yang, Y., Ren, Y., Zhu, Q., Feng, Z., Li, C., Lin, S., Wang, R., Luo, X., Zhang, B.: Wavemamba: Wavelet-driven mamba fusion for rgb-infrared object detection. ICCV (2025)

    2025

  73. [73]

    IEEE Trans

    Zhu, Y., Sun, X., Wang, M., Huang, H.: Multi-modal feature pyramid transformer for rgb-infrared object detection. IEEE Trans. Intell. Transport. Syst. (2023)

    2023