Pith. sign in

REVIEW 6 major objections 5 minor 2 cited by

MHAF-YOLO: Multi-Branch Heterogeneous Auxiliary Fusion YOLO for accurate object detection

T0 review · 6 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a new neck design, MAFPN, lets a 7.1M-parameter YOLO reach 48.9% AP on COCO, beating YOLO11-s by 1.9 points with 24.4% fewer parameters.

desk verdict Genuine incremental architecture with clean internal ablations; the SOTA claim over YOLO11 rests on uncontrolled baselines and val-set tuning. read the letter →

arxiv 2502.04656 v2 pith:J7EZUN37 submitted 2025-02-07 cs.CV

Zhiqiang Yang , Qiu Guan , Zhongwen Yu , Xinli Xu , Haixia Long , Sheng Lian , Haigen Hu , Ying Tang This is my paper · ORCID
classification cs.CV
keywords objectdetectionYOLOmulti-scalefeaturefusionpyramidnetworkreparameterizationlargekernelconvolutionmodelefficiencyCOCO
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 claims that the standard PAFPN neck used in most YOLO detectors wastes the information that shallow backbone layers carry about small objects, and that this can be fixed with a new neck architecture. The proposed MHAF-YOLO adds two auxiliary fusion paths, SAF at shallow layers and AAF at deep layers, plus a reparameterized heterogeneous multi-scale convolution module, and reports 48.9% AP on COCO at 7.1M parameters, 1.9 points above YOLO11-s with 24.4% fewer parameters. The authors also report gains when the neck is plugged into YOLOv8n and Faster R-CNN, and improved instance segmentation and rotated object detection. If correct, it means feature-fusion design is a currently underused axis for making real-time detectors both smaller and more accurate.

What carries the argument

The load-bearing object is MAFPN (Multi-Branch Auxiliary FPN), a neck that adds Superficial Assisted Fusion (SAF) and Advanced Assisted Fusion (AAF) to the standard bottom-up and top-down paths. SAF injects shallow backbone features into the neck's first fusion blocks to preserve small-object spatial detail; AAF densely connects four feature layers at the deeper output stages so the head receives richer gradients. Alongside it, RepHMS (Reparameterized Heterogeneous Multi-Scale) runs parallel small and large depthwise convolutions during training and folds them into a single kernel at inference, while GHFKS (Global Heterogeneous Flexible Kernel Selection) chooses larger kernels on higher-resolution layers to widen receptive fields. The mechanism's work is to make multi-scale information flow cheaply: no extra inference cost for the multi-kernel branches, and no extra parameters for the fusion paths beyond a few 1x1 convolutions.

What would settle it

Train MHAF-YOLO-s and YOLO11-s from scratch with the identical data pipeline, epochs, and optimizer on COCO train2017 and compare AP on COCO val2017; if the reported 1.9-point edge at 7.1M parameters does not replicate, the central advantage claim is not robust.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a one-stage detector's neck can be rebuilt to fuse non-adjacent scales in both directions, letting shallow spatial detail survive into deep layers and giving small objects a disproportionate accuracy gain. Specifically, MHAF-YOLO-s achieves 48.9% AP (49.1% with one-to-many NMS training) at 7.1M parameters and 25.3 GFLOPs on COCO val2017, which the authors report as state-of-the-art among real-time detectors and as a 1.9 AP improvement over YOLO11-s at 24.4% fewer parameters. The same architecture transfers to instance segmentation (39.7% seg AP for the small variant) and rotated object detection (81.10% mAP on DOTA-v1.0 in multi-scale mode), which the authors offer as evidence of generalization.

Load-bearing premise

The claim rests on the published baseline numbers being accurate and fairly comparable, and on COCO val2017 serving both to tune the design and to score the final result, so a baseline run with a different training recipe could change the reported gaps.

Editorial extensions

If this is right

  • Real-time object detectors can improve accuracy per parameter by redesigning the neck instead of scaling up the backbone or head.
  • MAFPN as a plug-in reports AP gains on other detectors: +1.7 on YOLOv8n (with fewer parameters and 200 fewer epochs) and +1.2 on Faster R-CNN.
  • RepHMS-style reparameterized heterogeneous kernels give multi-scale receptive fields at no additional inference cost, since training-time branches merge into a single convolution.
  • The same detector body transfers to instance segmentation and rotated object detection with gains over YOLO11 counterparts.

Reading between the lines

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

  • The ablations select hyperparameters on COCO val2017 and the final comparisons are also on val2017, so part of the reported gap may reflect tuning to that split; a test-dev evaluation would show whether the SOTA claim holds outside the selection set.
  • The authors state that inference speed trails YOLOv10 and YOLO11; if latency rather than parameters is the deployment constraint, the practical advantage over those models is smaller than the AP-per-parameter numbers suggest.
  • The training recipe differs from baselines (cached-mixup from RTMDet and lower-probability copy-paste), so a recipe-matched controlled comparison would isolate how much of the gain is architectural.
  • Since MAFPN is plug-and-play and RepHMS can enter any network, the same design is a natural test in transformer-based detectors, not just convolutional YOLO variants.
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

6 major / 5 minor

Summary. The paper proposes MHAF-YOLO, a YOLO-family detector built on three main architectural ingredients: a Multi-Branch Auxiliary FPN (MAFPN) neck with Superficial Assisted Fusion (SAF) and Advanced Assisted Fusion (AAF), a Reparameterized Heterogeneous Multi-Scale (RepHMS) feature-extraction block using reparameterized heterogeneous depthwise convolutions, and a Global Heterogeneous Flexible Kernel Selection (GHFKS) mechanism that assigns larger kernels to different feature scales. The authors report ablations on COCO val2017, comparisons against published real-time detector results on COCO and VOC, and transfer experiments to instance segmentation and rotated object detection. The central quantitative claim is that MHAF-YOLO-s reaches 48.9 AP at 7.1M parameters and 25.3 GFLOPs, surpassing YOLO11-s (47.0 AP, 9.4M) and YOLOv12-s (48.0 AP, 9.3M), and the paper concludes that MHAF-YOLO achieves state-of-the-art performance in real-time object detection.

Significance. If the empirical claims were established under a controlled protocol, this would be a practically valuable architecture paper. The internal ablations in Tables 2, 3, and 5 are systematic and show that each proposed component contributes positively under the authors' training recipe; the reparameterization identity in Eq. (6) is algebraic, and the code is publicly released. The transfer demonstrations to segmentation and rotated detection support the generality of the modules. However, the central SOTA claim currently rests on published baseline numbers from heterogeneous training protocols and on model selection and evaluation using the same COCO val2017 split, so the magnitude of the architectural gain over YOLO11/YOLOv12 is not yet established. The paper also never reports inference latency while calling itself real-time.

major comments (6)
  1. [§4.2, Table 6] The main SOTA claim is measured against published baseline numbers rather than models retrained under the paper's protocol. MHAF-YOLO is trained from scratch for 500 epochs with SGD, cached-mixup, and low-probability copy-paste, while YOLOv10, YOLO11, YOLOv12, Gold-YOLO, and MAF-YOLO results in Table 6 come from their own schedules and augmentation stacks; the 1.9 AP gap over YOLO11-s and 0.9 AP gap over YOLOv12-s therefore conflate architecture with training recipe. Please retrain at least YOLOv10-s, YOLO11-s, YOLOv12-s, Gold-YOLO-s, and MAF-YOLO-s under the identical 500-epoch recipe, or report official and same-recipe numbers side by side, and give full training hyperparameters such as batch size, input resolution, learning-rate schedule, and augmentation probabilities.
  2. [§4.3–§4.6, Table 6] The experimental protocol selects architecture hyperparameters on COCO val2017 in Sections 4.3–4.5 (RepHMS depths, GHFKS kernel sizes, SAF/AAF configurations) and then evaluates the final models on the same val2017 split in Section 4.6, Table 6. This selection-on-evaluation procedure can inflate both the ablations and the SOTA comparison, and the paper reports only APval with no COCO test-dev numbers. Please report test-dev AP, or make an explicit train/val separation for model selection and final evaluation, and state which reported numbers were used for architecture decisions.
  3. [§5 (Conclusions), §1, Fig. 2] The paper repeatedly frames MHAF-YOLO as a real-time detector, but no latency or FPS measurement appears anywhere, and the conclusion explicitly concedes that inference speed still lags behind cutting-edge models such as YOLOv10 and YOLO11 because of MAFPN complexity and large-kernel depthwise convolutions. The real-time claim is therefore unsupported by the evidence; please add latency measurements on identical hardware, batch size, and input resolution for all compared models and revise the SOTA claims to separate accuracy/parameter trade-off from speed.
  4. [§4.6.1, Table 6] The nano-scale result is internally inconsistent with the claim that MHAF-YOLO surpasses existing real-time detectors: MHAF-YOLO-n reaches 42.3 AP, while the authors' own MAF-YOLO-n in the same table reaches 42.4 AP despite having more parameters. If MHAF-YOLO is meant to improve on MAF-YOLO, the table should be explained in terms of training recipe or capacity target; otherwise the claim of uniform superiority across N/S/M scales should be qualified.
  5. [§4.4.2, Table 4] The plug-and-play claim for MAFPN is partly confounded in the YOLOv8n experiment: YOLOv8n-MAFPN is trained for 300 epochs while YOLOv8n-PAFPN is trained for 500 epochs, and its channel count is changed to make the model smaller, so the +1.7 AP gain cannot be attributed solely to the neck. Please provide an equal-epoch, equal-capacity comparison for YOLOv8n and specify the training schedule, resolution, and augmentation used for the Faster R-CNN comparison.
  6. [§4.2, Table 6] Section 4.2 states that all scales of MHAF-YOLO are trained from scratch without relying on large-scale datasets such as ImageNet or pre-trained weights, but Table 6 includes rows MHAF-YOLO-n* and MHAF-YOLO-s* whose footnote says they are trained with a pretrained backbone. The text never explains how these starred models are pretrained or why they are included, and the table presents their higher numbers alongside the from-scratch results. Please remove the starred rows or describe the pretraining protocol explicitly, and keep the from-scratch claim consistent with the tables.
minor comments (5)
  1. [Table 6, References] YOLO11-n and YOLO11-s are cited as reference [16] in Table 6, but reference [16] is the authors' MAF-YOLO paper; YOLO11 should be cited as [18] as in Table 7, otherwise the provenance of the baseline numbers is unclear.
  2. [§4.2] The implementation details list the optimizer, hardware, and epoch count but omit batch size, input resolution, learning rate, momentum/weight decay, and the actual cached-mixup/copy-paste probabilities; these details are needed to reproduce the 500-epoch runs and to verify the claimed parameter-accuracy trade-offs.
  3. [Eq. (6)] The notation in Eq. (6) is ambiguous: the kernel indices K_{2n-1} and K_{2n-(2i+1)} are not defined precisely, and the step of zero-padding smaller depthwise kernels before summation should be stated explicitly, since the correctness of the reparameterization merge relies on that padding.
  4. [§4.3.2, Table 2] The phrase saying that the Cascade strategy achieves 'lossless performance improvement' is inaccurate if 'lossless' refers to accuracy, because adding Cascade changes AP from 41.9 to 42.3 in Table 2; if the intended meaning is 'no extra parameters/FLOPs', the wording should say so.
  5. [Table 9, §4.6.4] Table 9 contains the label 'YOLOv11-n-obb' while the model is YOLO11-n-obb, and the rotated-detection section does not state the exact crop overlap and test-time augmentation settings for the MHAF rows beyond the default description; please align notation and protocol descriptions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reparameterization identity is algebraic and the SOTA claims are anchored to external baselines; MAF-YOLO self-citations are not load-bearing.

full rationale

The paper's derivation chain does not reduce to its own inputs. Equation (6) is a direct algebraic identity: BN-folded small depthwise kernels are zero-padded and summed to form a large kernel, with no fitted parameter or target result assumed. The ablations in Tables 1-5 are incremental, controlled changes from a YOLOv10n baseline with AP measured on COCO; none of the reported gains is a fitted quantity renamed as a prediction. The main SOTA claims in Table 6 are comparisons against externally published detectors (YOLO11, YOLOv12, Gold-YOLO, RTMDet, etc.), so the central claim does not depend on the authors' prior MAF-YOLO paper. MAF-YOLO [16] appears as a baseline row and as the RepHELAN block in Table 1, but those citations are not load-bearing: removing them would not change the YOLO11/YOLOv12 comparisons or the internal ablations. The use of COCO val2017 for both ablations and final reporting, and the reliance on published baseline numbers, are experimental-comparability and model-selection concerns rather than definitional circularity, and no equation-level reduction from output to input is exhibited. Therefore no circular steps are identified.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical, so the ledger is mostly design hyperparameters and benchmark assumptions rather than mathematical axioms. The reparameterization identity is standard algebra. The main burden is that the architecture is tuned on the same COCO val split used for final numbers, and that SOTA comparisons inherit unverified baseline numbers.

free parameters (6)
  • RepHMS branch count N and block depth M = not stated; default used in experiments
    Control the width/depth of the multi-branch block and are tuned via ablations on COCO val; they directly affect the 2.2M/7.2G nano configuration.
  • GHFKS kernel-size schedule = 3,5,7,9 in backbone; 5,7,9 in MAFPN
    Chosen by hand as a global schedule across resolutions; no adaptive mechanism is defined, and the choices are validated on COCO val.
  • SAF shallow-channel ratio = half the deep-layer channels
    Selected after experiments; the paper states this ratio avoids disturbing subsequent learning.
  • AAF channel equalization = equal channels across layers
    Chosen after experiments showed half-channel settings degraded performance; based on the postulate in Section 3.3.2.
  • RepHConv kernel combination = large kernel (e.g., 7x7) plus small kernels (3x3, 5x5)
    Parallel kernels are merged at inference; the exact set is a design choice tested through the RepHConv ablation row.
  • Training augmentation and schedule = 500 epochs, cached-mixup, lower-probability copy-paste, 10-epoch decay off
    Borrowed from YOLOv10/RTMDet and adjusted for this model; not independently justified.
assumptions (5)
  • standard math Addition of zero-padded depthwise convolution kernels after BN fusion yields an equivalent single kernel.
    Used in Eq. (6) for RepHDWConv reparameterization; this is standard algebraic equivalence for convolutions.
  • domain assumption Larger receptive fields improve detection, while smaller receptive fields are preferable for small objects.
    Borrowed from Trident Network and YOLO-MS in Section 2.3 and 3.2; used to justify GHFKS and RepHConv.
  • domain assumption Published baseline results in Tables 6-9 are accurate and were obtained under sufficiently comparable conditions.
    The SOTA comparisons use numbers from other papers without re-running them under identical training settings.
  • domain assumption COCO val2017 is an acceptable target for both model selection and final reporting.
    Ablations in Sections 4.3-4.5 select modules on val2017, and Section 4.6 reports final results on the same split.
  • ad hoc to paper AAF channel equalization is valid because initial guiding information is already embedded in shallow MAFPN layers.
    Stated as a postulate in Section 3.3.2; used to set equal channel counts in the AAF fusion equation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MHAF-YOLO: Multi-Branch Heterogeneous Auxiliary Fusion YOLO for accurate object detection." pith.science (2026). https://pith.science/paper/J7EZUN37

@misc{pith2026250204656,
  author       = {Pith},
  title        = {Pith review of: MHAF-YOLO: Multi-Branch Heterogeneous Auxiliary Fusion YOLO for accurate object detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J7EZUN37}},
  note         = {Machine review of arXiv:2502.04656}
}
read the original abstract

Due to the effective multi-scale feature fusion capabilities of the Path Aggregation FPN (PAFPN), it has become a widely adopted component in YOLO-based detectors. However, PAFPN struggles to integrate high-level semantic cues with low-level spatial details, limiting its performance in real-world applications, especially with significant scale variations. In this paper, we propose MHAF-YOLO, a novel detection framework featuring a versatile neck design called the Multi-Branch Auxiliary FPN (MAFPN), which consists of two key modules: the Superficial Assisted Fusion (SAF) and Advanced Assisted Fusion (AAF). The SAF bridges the backbone and the neck by fusing shallow features, effectively transferring crucial low-level spatial information with high fidelity. Meanwhile, the AAF integrates multi-scale feature information at deeper neck layers, delivering richer gradient information to the output layer and further enhancing the model learning capacity. To complement MAFPN, we introduce the Global Heterogeneous Flexible Kernel Selection (GHFKS) mechanism and the Reparameterized Heterogeneous Multi-Scale (RepHMS) module to enhance feature fusion. RepHMS is globally integrated into the network, utilizing GHFKS to select larger convolutional kernels for various feature layers, expanding the vertical receptive field and capturing contextual information across spatial hierarchies. Locally, it optimizes convolution by processing both large and small kernels within the same layer, broadening the lateral receptive field and preserving crucial details for detecting smaller targets. The source code of this work is available at: https://github.com/yang-0201/MHAF-YOLO.

Figures

Figures reproduced from arXiv: 2502.04656 by the authors.

Figure 1
Figure 1. Diagram (a) illustrates the PAFPN structure in YOLOv10, while diagrams (b) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Comparisons with other state-of-the-art real-time object detectors in terms of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Different feature extraction blocks. (a)The basic block of CSPNet in [5, 12, [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Overview of the network architecture of MHAF-YOLO. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: The architecture of Superficial Assisted Fusion. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: The architecture of Advanced Assisted Fusion. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: The architecture of RepHMS. After designing the MAFPN structure in the preceding section, another challenge lies in efficiently designing the feature extraction block within the entire architecture. This section presents the design of a powerful encoder architecture th…
Figure 8
Figure 8. Figure 8: Let µ, σ, γ, β as the accumulated mean, standard deviation and learned scaling factor and bias of the BN layer. RepHDW(x) represents a RepHDWConv parameters. I indicates input feature maps, Kn and Bn show the weight and bias of the convolution with a n × n kernel. Firs…
Figure 8
Figure 8. Figure 8: An example of re-parameterizing a small kernel (e.g., 3 [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Multi-scale performance comparison chart. [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]
Figure 10
Figure 10. Figure 10: The detection results of YOLOv9n, YOLOv10n, YOLO11n, MHAF-YOLOn. [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. CollabOD: Collaborative Multi-Backbone with Cross-scale Vision for UAV Small Object Detection

    cs.CV 2026-03 conditional novelty 4.0 of 10

    CollabOD improves UAV small-object detection via dual-path detail preservation, dense aggregation, bilateral reweighting, and a reparameterized detail-aware head, reporting 52.4 AP50 on VisDrone at 65.5 GFLOPs.

  2. MambaNeXt-YOLO: A Hybrid State Space Model for Real-time Object Detection

    cs.CV 2025-06 conditional novelty 4.0 of 10

    MambaNeXt-YOLO, a hybrid CNN-Mamba detector, reports 66.6% mAP on PASCAL VOC from scratch and 27.8% on DOTA v1.5, with marginal gains over Mamba-YOLO and lower frame rates than several lightweight YOLO variants.

Reference graph

Works this paper leans on

51 extracted references · 28 canonical work pages · cited by 2 Pith papers

  1. [1]

    Redmon, You only look once: Unified, real-time object detection, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016

    J. Redmon, You only look once: Unified, real-time object detection, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016

    work page 2016

  2. [2]

    Redmon, A

    J. Redmon, A. Farhadi, Yolo9000: better, faster, stronger, in: Proceed- ings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 7263–7271

    2017

  3. [3]

    Redmon, Yolov3: An incremental improvement, arXiv preprint arXiv:1804.02767 (2018)

    J. Redmon, Yolov3: An incremental improvement, arXiv preprint arXiv:1804.02767 (2018)

    arXiv 2018

  4. [4]

    Bochkovskiy, C.-Y

    A. Bochkovskiy, C.-Y. Wang, H.-Y. M. Liao, Yolov4: Optimal speed and accuracy of object detection, arXiv preprint arXiv:2004.10934 (2020)

    arXiv 2020

  5. [5]

    Jocher, A

    G. Jocher, A. Chaurasia, A. Stoken, et al, ultralytics/yolov5: v7. 0- yolov5 sota realtime instance segmentation, Zenodo (2022)

    work page 2022

  6. [6]

    Z. Ge, S. Liu, F. Wang, Z. Li, J. Sun, Yolox: Exceeding yolo series in 2021, arXiv preprint arXiv:2107.08430 (2021)

    arXiv 2021

  7. [7]

    C. Li, L. Li, Y. Geng, et al, Yolov6 v3. 0: A full-scale reloading, arXiv preprint arXiv:2301.05586 (2023)

    arXiv 2023

  8. [8]

    C.-Y. Wang, A. Bochkovskiy, H.-Y. M. Liao, Yolov7: Trainable bag-of- freebies sets new state-of-the-art for real-time object detectors, in: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 7464–7475

    work page 2023

Show all 51 references
  1. [9]

    Jocher, A

    G. Jocher, A. Chaurasia, J. Qiu, Yolo by ultralytics, URL: https://github.com/ultralytics/ultralytics/tree/v8.2.0 (2023)

    2023

  2. [10]

    X. Xu, Y. Jiang, W. Chen, et al, Damo-yolo: A report on real-time object detection design, arXiv preprint arXiv:2211.15444 (2022)

    2022 arXiv

  3. [11]

    C. Lyu, W. Zhang, H. Huang, et al, Rtmdet: An empirical study of designing real-time object detectors, arXiv preprint arXiv:2212.07784 (2022). 30

    2022 arXiv

  4. [12]

    S. Xu, X. Wang, W. Lv, Q. Chang, C. Cui, K. Deng, G. Wang, Q. Dang, S. Wei, Y. Du, et al., Pp-yoloe: An evolved version of yolo, arXiv preprint arXiv:2203.16250 (2022)

    2022 arXiv

  5. [13]

    C. Wang, W. He, Y. Nie, et al, Gold-yolo: Efficient object detector via gather-and-distribute mechanism, arXiv preprint arXiv:2309.11331 (2023)

    2023 arXiv

  6. [14]

    Y. Chen, X. Yuan, R. Wu, et al, Yolo-ms: Rethinking multi-scale representation learning for real-time object detection, arXiv preprint arXiv:2308.05480 (2023)

    2023 arXiv

  7. [15]

    Wang, I.-H

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

    2024 arXiv

  8. [16]

    Z. Yang, Q. Guan, K. Zhao, J. Yang, X. Xu, H. Long, Y. Tang, Multi-branch auxiliary fusion yolo with re-parameterization hetero- geneous convolutional for accurate object detection, arXiv preprint arXiv:2407.04381 (2024)

    2024 arXiv

  9. [17]

    A. Wang, H. Chen, L. Liu, K. Chen, Z. Lin, J. Han, G. Ding, Yolov10: Real-time end-to-end object detection, arXiv preprint arXiv:2405.14458 (2024)

    2024 arXiv

  10. [18]

    Jocher, A

    G. Jocher, A. Chaurasia, J. Qiu, Yolo by ultralytics, URL: https://github.com/ultralytics/ultralytics (2024)

    2024

  11. [19]

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

    2025 arXiv

  12. [20]

    T.-Y. Lin, P. Doll´ ar, R. Girshick, et al, Feature pyramid networks for object detection, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 2117–2125

    2017

  13. [21]

    K. Wang, J. H. Liew, Y. Zou, D. Zhou, J. Feng, Panet: Few-shot image semantic segmentation with prototype alignment, in: proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 9197–9206. 31

    2019

  14. [22]

    Wang, H.-Y

    C.-Y. Wang, H.-Y. M. Liao, Y.-H. Wu, P.-Y. Chen, J.-W. Hsieh, I.-H. Yeh, Cspnet: A new backbone that can enhance learning capability of cnn, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 2020, pp. 390–391

    2020

  15. [23]

    C. Feng, Y. Zhong, Y. Gao, M. R. Scott, W. Huang, Tood: Task-aligned one-stage object detection, in: 2021 IEEE/CVF International Confer- ence on Computer Vision (ICCV), IEEE Computer Society, 2021, pp. 3490–3499

    2021

  16. [24]

    Z. Sun, M. Lin, X. Sun, Z. Tan, H. Li, R. Jin, Mae-det: Revisiting max- imum entropy principle in zero-shot nas for efficient object detection, arXiv preprint arXiv:2111.13336 (2021)

    2021 arXiv

  17. [25]

    G. Yang, J. Lei, Z. Zhu, S. Cheng, Z. Feng, R. Liang, Afpn: Asymptotic feature pyramid network for object detection, in: 2023 IEEE Interna- tional Conference on Systems, Man, and Cybernetics (SMC), IEEE, 2023, pp. 2184–2189

    2023

  18. [26]

    Y. Li, Y. Chen, N. Wang, Z. Zhang, Scale-aware trident networks for object detection, in: Proceedings of the IEEE/CVF international con- ference on computer vision, 2019, pp. 6054–6063

    2019

  19. [27]

    X. Ding, X. Zhang, N. Ma, et al, Repvgg: Making vgg-style convnets great again, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 13733–13742

    2021

  20. [28]

    X. Ding, X. Zhang, J. Han, G. Ding, Scaling up your kernels to 31x31: Revisiting large kernel design in cnns, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11963– 11975

    2022

  21. [29]

    X. Ding, Y. Zhang, Y. Ge, S. Zhao, L. Song, X. Yue, Y. Shan, Unire- plknet: A universal perception large-kernel convnet for audio video point cloud time-series and image recognition, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, p...

    2024

  22. [30]

    T.-Y. Lin, M. Maire, S. Belongie, et al, Microsoft coco: Common objects in context, in: Computer Vision–ECCV 2014: 13th European Confer- 32 ence, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, Springer, 2014, pp. 740–755

    2014

  23. [31]

    Everingham, S

    M. Everingham, S. A. Eslami, L. Van Gool, C. K. Williams, J. Winn, A. Zisserman, The pascal visual object classes challenge: A retrospec- tive, International journal of computer vision 111 (2015) 98–136

    2015

  24. [32]

    G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, L. Zhang, Dota: A large-scale dataset for object detection in aerial images, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3974–3983

    2018

  25. [33]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, L. Fei-Fei, Imagenet: A large-scale hierarchical image database, in: 2009 IEEE conference on computer vision and pattern recognition, Ieee, 2009, pp. 248–255

    2009

  26. [34]

    Ghiasi, Y

    G. Ghiasi, Y. Cui, A. Srinivas, R. Qian, T.-Y. Lin, E. D. Cubuk, Q. V. Le, B. Zoph, Simple copy-paste is a strong data augmentation method for instance segmentation, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 2918–2928

    2021

  27. [35]

    S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: Towards real-time object detection with region proposal networks, IEEE transactions on pattern analysis and machine intelligence 39 (6) (2016) 1137–1149

    2016

  28. [36]

    M. Tan, R. Pang, Q. V. Le, Efficientdet: Scalable and efficient object detection, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 10781–10790

    2020

  29. [37]

    W. Lv, Y. Zhao, Q. Chang, K. Huang, G. Wang, Y. Liu, Rt-detrv2: Im- proved baseline with bag-of-freebies for real-time detection transformer, arXiv preprint arXiv:2407.17140 (2024)

    2024 arXiv

  30. [38]

    Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, S. Lin, B. Guo, Swin transformer: Hierarchical vision transformer using shifted windows, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10012–10022

    2021

  31. [39]

    Z. Cai, N. Vasconcelos, Cascade r-cnn: High quality object detection and instance segmentation, IEEE transactions on pattern analysis and machine intelligence 43 (5) (2019) 1483–1498. 33

    2019

  32. [40]

    Z. Liu, H. Mao, C.-Y. Wu, C. Feichtenhofer, T. Darrell, S. Xie, A con- vnet for the 2020s, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11976–11986

    2022

  33. [41]

    X. Zhu, W. Su, L. Lu, B. Li, X. Wang, J. Dai, Deformable detr: De- formable transformers for end-to-end object detection, arXiv preprint arXiv:2010.04159 (2020)

    2020 arXiv

  34. [42]

    Y. Li, H. Mao, R. Girshick, K. He, Exploring plain vision transformer backbones for object detection, in: European conference on computer vision, Springer, 2022, pp. 280–296

    2022

  35. [43]

    Zhang, F

    H. Zhang, F. Li, S. Liu, et al, Dino: Detr with improved denoising anchor boxes for end-to-end object detection, arXiv preprint arXiv:2203.03605 (2022)

    2022 arXiv

  36. [44]

    K. Chen, J. Pang, J. Wang, Y. Xiong, X. Li, S. Sun, W. Feng, Z. Liu, J. Shi, W. Ouyang, et al., Hybrid task cascade for instance segmentation, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4974–4983

    2019

  37. [45]

    X. Wang, R. Zhang, T. Kong, L. Li, C. Shen, Solov2: Dynamic and fast instance segmentation, Advances in Neural information processing systems 33 (2020) 17721–17732

    2020

  38. [46]

    X. Xie, G. Cheng, J. Wang, X. Yao, J. Han, Oriented r-cnn for object detection, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 3520–3529

    2021

  39. [47]

    X. Cai, Q. Lai, Y. Wang, W. Wang, Z. Sun, Y. Yao, Poly kernel inception network for remote sensing detection, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 27706–27716

    2024

  40. [48]

    X. Yang, Y. Zhou, G. Zhang, J. Yang, W. Wang, J. Yan, X. Zhang, Q. Tian, The kfiou loss for rotated object detection, arXiv preprint arXiv:2201.12558 (2022)

    2022 arXiv

  41. [49]

    D. Wang, Q. Zhang, Y. Xu, J. Zhang, B. Du, D. Tao, L. Zhang, Advanc- ing plain vision transformer toward remote sensing foundation model, IEEE Transactions on Geoscience and Remote Sensing 61 (2022) 1–15. 34

    2022

  42. [50]

    Zhang, Y

    Q. Zhang, Y. Xu, J. Zhang, D. Tao, Vitaev2: Vision transformer ad- vanced by exploring inductive bias for image recognition and beyond, International Journal of Computer Vision 131 (5) (2023) 1141–1162

    2023

  43. [51]

    Y. Li, X. Li, Y. Dai, Q. Hou, L. Liu, Y. Liu, M.-M. Cheng, J. Yang, Lsknet: A foundation lightweight backbone for remote sensing, Interna- tional Journal of Computer Vision (2024) 1–22. 35

    2024

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.