Pith. sign in

REVIEW 5 major objections 4 minor 48 references

YOLO-FDA: Integrating Hierarchical Attention and Detail Enhancement for Surface Defect Detection

T0 review · 5 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that YOLO-FDA, a YOLOv5 variant with a BiFPN neck, directional-detail fusion, and two attention fusion modules, improves surface-defect detection accuracy on steel and synthetic-texture benchmarks, reporting 74.9% mAP50…

desk verdict Routine YOLOv5 variant with a real controlled-baseline gain, but SOTA and robustness claims outrun the evidence. read the letter →

arxiv 2506.21135 v1 pith:Y2KRYBIY submitted 2025-06-26 cs.CV

Jiawei Hu This is my paper
classification cs.CV
keywords surfacedefectdetectionYOLOv5BiFPNasymmetricconvolutionattentionfusionGC10-DETDAGM2007
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper targets industrial surface defect detection, where defects are small, elongated, and embedded in textured backgrounds. It claims that standard YOLO-style feature fusion drops the fine directional detail these defects need, and that a modified YOLOv5, named YOLO-FDA, fixes this by combining a BiFPN neck, a Detail-directional Fusion Module that uses asymmetric 1x7 and 7x1 convolutions, and two attention-based fusion strategies. On the GC10-DET steel dataset it reports 74.9% mAP50, 4.6 points above its own re-run YOLOv5s baseline, and on DAGM2007 it reports 67.9% mAP50-95, 2.7 points above the baseline. The paper's conclusion is that this combination improves both accuracy and robustness across defect types and scales.

What carries the argument

The central mechanism is the network-neck redesign: a BiFPN-style bidirectional multi-scale fusion; the Detail-directional Fusion Module (DDFM), which fuses the second-lowest and fourth backbone layers and applies 1x7 and 7x1 asymmetric convolutions to expose directional texture; and two learnable fusion rules at the three-path merge nodes, namely Attention-weighted Concatenation (AC), which weights each feature map before concatenation, and Cross-layer Attention Fusion (CAF), which computes a softmax-weighted sum across layers. Together these modules carry the claim that fine detail can be preserved early and redundant cross-scale information can be suppressed late.

What would settle it

Re-run every compared detector, including DSL-YOLO, YOLOv9s, and the multi-layer fusion variant, on GC10-DET and DAGM2007 with exactly the paper's settings of 250 epochs, batch size 8, and the same data split, and check whether YOLO-FDA still leads in mAP50 and mAP50-95; if the margin shrinks to noise or reverses, the claimed state-of-the-art advantage does not hold. A second check is whether removing DDFM while keeping AC and CAF changes the score, since the paper's own ablation shows DDFM alone lowers mAP.

Watch

Extended reading notes

Core claim

YOLO-FDA claims that replacing YOLOv5's PANet neck with BiFPN and adding targeted detail and attention modules yields consistent gains on surface-defect benchmarks. The Detail-directional Fusion Module enriches the second-lowest feature layer by upsampling the fourth backbone layer, concatenating it with low-level features, and adding two asymmetric convolutions (1x7 horizontal, 7x1 vertical) so elongated and directional defects are not lost. Attention-weighted Concatenation scales each input feature map before concatenation to preserve all channels, while Cross-layer Attention Fusion learns a softmax-weighted sum across layers to suppress redundancy. The reported results are 74.9% mAP50 on GC10-DET and 67.9% mAP50-95 on DAGM2007, with ablation experiments showing the full combination outperforms any single module.

Load-bearing premise

The load-bearing premise is that published state-of-the-art numbers trained with different epochs, batch sizes, and GPUs can be compared directly against the paper's own runs, since only the YOLOv5s baseline was trained under identical conditions.

Editorial extensions

If this is right

  • If the reported gains hold, industrial inspection systems can adopt YOLO-FDA directly on top of existing YOLOv5 deployments for a 4.6-point mAP50 gain on steel surfaces without changing the detection head.
  • The AC and CAF fusion rules imply that preserving channel diversity early and mixing additively late improves small-defect accuracy in feature-concatenation networks.
  • Asymmetric 1x7 and 7x1 convolutions provide a concrete recipe for detecting defects with high aspect ratio, such as welds and inclusions.
  • On DAGM2007, the C1 category gain (84.5% versus 75.9% for YOLOv5s) suggests the module combination particularly helps defect classes with lower baseline detectability.
  • The consistent gains on two different benchmarks indicate the neck-level changes are portable across data domains rather than tuned to one texture type.

Reading between the lines

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

  • The paper's ablation shows DDFM by itself reduces mAP; an editorial inference is that the net gain is carried by AC and CAF, so a simpler model with only attention fusion might match YOLO-FDA on these benchmarks.
  • Because most comparison numbers come from other papers with different training epochs, batch sizes, and GPUs, the true rank against DSL-YOLO and YOLOv9s may change under uniform training conditions.
  • The same two-stage fusion recipe could transfer to other one-stage detectors handling elongated objects, such as PCB defects or fabric flaw detection, though this is beyond the paper's evidence.
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

5 major / 4 minor

Summary. The paper proposes YOLO-FDA, a YOLOv5-based detector for surface defect detection that replaces the PANet neck with a BiFPN-style architecture, adds a Detail-directional Fusion Module (DDFM) using asymmetric convolutions on a low-level feature path, and introduces two attention-based fusion mechanisms (AC and CAF) at three-path fusion nodes. Experiments are reported on GC10-DET and DAGM2007. The authors claim consistent state-of-the-art accuracy and robustness, with the main quantitative results being 74.9% mAP50 on GC10-DET (4.6 points over their own YOLOv5s baseline) and 67.9% mAP50-95 on DAGM2007 (2.7 points over YOLOv5s). Ablation studies on GC10-DET show that BiFPN helps, the addition of DDFM alone slightly hurts, and the AC/CAF modules improve results, with the full model giving the best performance.

Significance. If the empirical claims are taken at face value, the work is an incremental but potentially useful engineering contribution to YOLO-based surface defect detection. The controlled baseline improvement of 4.6 mAP50 over a re-run YOLOv5s is a genuine positive result, and the combination of directional detail enhancement with attention-weighted fusion is a reasonable design direction. However, the headline claim of consistent superiority over prior state-of-the-art is not supported by the evidence as presented: almost all SOTA comparisons are quoted from papers with different training protocols, and the robustness claim has no dedicated experiments. The novelty is moderate, and the paper does not provide code, model weights, or a fully specified architecture, which limits reproducibility and immediate practical uptake.

major comments (5)
  1. [§4.4, Table 1] The central claim of consistent outperformance over existing state-of-the-art rests on comparisons that are not matched in training protocol. The paper itself acknowledges that YOLOv3, YOLOv7tiny, YOLOv5n, YOLOv8n, YOLOv7, and DSL-YOLO are taken from related work [40] with 200 training epochs and a different GPU, and that YOLOv9s and YOLOv5s-improved come from [41] with 500 epochs and batch size 32, whereas the authors use 250 epochs and batch size 8. The margin over DSL-YOLO (72.5 vs. 74.9 mAP50) could plausibly shrink or reverse under identical training budgets. To support the SOTA claim, the authors should re-run the key competitors under their own protocol or, failing that, weaken the claim to 'improvement over the YOLOv5s baseline' and clearly label quoted numbers as incomparable.
  2. [Abstract and §4.4] The abstract and conclusion claim robustness, but no experiment in the paper perturbs scale, noise, lighting, texture, or any other input condition. The visualization section shows only a few qualitative examples. The term 'robustness' is used without an operational definition or quantitative evaluation. The authors should either add a robustness evaluation (e.g., noise injection, scaling, or contrast perturbations) or remove the robustness claim from the abstract and conclusions.
  3. [§4.5, Fig. 5] The ablation shows that adding DDFM to BiFPN decreases mAP50 compared with BiFPN alone, so the sole detail-enhancement module contributes positively only when combined with the AC/CAF attention modules. The paper's explanation that simple concatenation causes redundancy is speculative and not supported by the ablation design. This is a load-bearing point because DDFM is one of the paper's three named contributions. The authors should provide a more thorough analysis, e.g., ablating DDFM with each attention module separately, or reporting intermediate feature statistics, to clarify whether DDFM has any independent value.
  4. [Eqs. (4) and (9), §4.3] The reduction ratio r that defines the bottleneck channels C' = C/r in the AC module (Eq. 4) and the CAF module (Eq. 9) is never given a value. Without this hyperparameter, the architecture is not fully specified and the experiments cannot be exactly reproduced. The paper should state the value of r used in all experiments (and, if varied, report sensitivity).
  5. [Tables 1 and 2] No variance estimates are reported for any of the runs, and each configuration appears to be trained once. On DAGM2007, the gain over the same-environment YOLOv9 baseline is only 0.8 mAP50-95 (67.9 vs. 67.1), which is likely within the run-to-run variation of a single-seed training. The authors should report results over multiple seeds (at least 3) with mean and standard deviation for at least their own model and the same-environment baselines, so the reader can judge whether the improvements are statistically meaningful.
minor comments (4)
  1. [§4.2] The evaluation metrics sentence says 'the Precision, Recall, and mAP are used to quantitatively evaluate the detection performance on the GC10-DET and DAGM2007, respectively,' which is grammatically unclear; it should specify that mAP50 is used for GC10-DET and mAP50-95 for DAGM2007 in the corresponding tables.
  2. [Fig. 1] The figure caption simply says 'Examples of surface defect from DAGM2007 and GC10-DET,' but the text refers to colored boxes in the second picture; the figure should label the defect categories (inclusion, weld, etc.) directly or the caption should explain the colored boxes.
  3. [Throughout] There are several typographical and formatting issues, including 'BiPFN' for BiFPN in the ablation discussion, 'indicator' used loosely for 'metric,' and incomplete reference entries (e.g., references [22], [33], [35] have no full bibliographic details). These should be cleaned up.
  4. [§5, Conclusion] The conclusion repeats the claim of 'multiscale robustness' without qualifications; given the absence of robustness experiments, this should be rephrased to reflect what is actually demonstrated.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: YOLO-FDA's claims are empirical and its ablations are reported honestly, including a module that reduces performance.

full rationale

The paper makes no claim of deriving a prediction from fitted inputs; YOLO-FDA is an architecture proposal evaluated by controlled experiments. The only same-protocol baseline, YOLOv5s, is re-run under the paper's own settings, and the module ablations in Section 4.5 are reported with concrete deltas, including Experiment 3 where DDFM alone decreases mAP50 and mAP50-95. That honesty is incompatible with fitted-input-called-prediction circularity. The SOTA comparisons in Tables 1 and 2 quote numbers from other papers with different training epochs, batch sizes, and GPUs; the paper explicitly flags this in Section 4.4, noting it is "slightly different... in terms of GPU type, etc." and that training rounds were 200 or 500 versus the paper's 250. This is a comparability or correctness limitation, not a definitional or self-referential reduction. The author is a single author, Jiawei Hu, and the cited related works by F. Shen et al. are not self-citations of this author, so no self-citation chain is load-bearing. The abstract's robustness claim is not backed by a dedicated robustness experiment, and the reduction ratio r in Eqs. (4) and (9) is left unspecified; both are reproducibility or support concerns, not circularity. Because no step in the claimed derivation reduces to its own input by construction, the circularity score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on empirical comparisons that assume cross-paper results are comparable, plus an unstated reduction ratio. No new physical or mathematical entities are introduced.

free parameters (1)
  • reduction ratio r in AC and CAF
    Used in Equations 4 and 9 to shrink channel dimension; value is never reported, so it must have been chosen by hand or via validation.
assumptions (3)
  • domain assumption DAGM2007 ellipse annotations can be converted to bounding boxes for detection
    Section 4.1 states this conversion; it can shift results relative to prior work that may use different evaluation conventions.
  • domain assumption Baseline results from prior papers are comparable despite different epochs, batch sizes, and hardware
    Section 4.4 explicitly notes these differences but still uses the copied numbers as direct comparisons.
  • domain assumption The proposed modules do not change the training loss or inference cost enough to break the comparison
    No FLOPs, latency, or memory measurements are provided, so the 'robustness' claim is only about accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of YOLO-FDA: Integrating Hierarchical Attention and Detail Enhancement for Surface Defect Detection." pith.science (2026). https://pith.science/paper/Y2KRYBIY

@misc{pith2026250621135,
  author       = {Pith},
  title        = {Pith review of: YOLO-FDA: Integrating Hierarchical Attention and Detail Enhancement for Surface Defect Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y2KRYBIY}},
  note         = {Machine review of arXiv:2506.21135}
}
read the original abstract

Surface defect detection in industrial scenarios is both crucial and technically demanding due to the wide variability in defect types, irregular shapes and sizes, fine-grained requirements, and complex material textures. Although recent advances in AI-based detectors have improved performance, existing methods often suffer from redundant features, limited detail sensitivity, and weak robustness under multiscale conditions. To address these challenges, we propose YOLO-FDA, a novel YOLO-based detection framework that integrates fine-grained detail enhancement and attention-guided feature fusion. Specifically, we adopt a BiFPN-style architecture to strengthen bidirectional multilevel feature aggregation within the YOLOv5 backbone. To better capture fine structural changes, we introduce a Detail-directional Fusion Module (DDFM) that introduces a directional asymmetric convolution in the second-lowest layer to enrich spatial details and fuses the second-lowest layer with low-level features to enhance semantic consistency. Furthermore, we propose two novel attention-based fusion strategies, Attention-weighted Concatenation (AC) and Cross-layer Attention Fusion (CAF) to improve contextual representation and reduce feature noise. Extensive experiments on benchmark datasets demonstrate that YOLO-FDA consistently outperforms existing state-of-the-art methods in terms of both accuracy and robustness across diverse types of defects and scales.

Figures

Figures reproduced from arXiv: 2506.21135 by the authors.

Figure 1
Figure 1. Examples of surface defect from DAGM2007 and GC10-DET. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall architecture of the proposed YOLO-FDA model. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the proposed Attention-weighted Concatenation module. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of the proposed Cross-layer Attention Fusion module. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Quantitative results of different ablation models tested on the GC10-DET [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Visualization results. Compared our method with baseline. The cyan frame is a [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 36 canonical work pages

  1. [40]

    Research on a metal surface defect detection algorithm based on dsl-yolo,

    Z. Wang, L. Zhao, H. Li, X. Xue, and H. Liu, “Research on a metal surface defect detection algorithm based on dsl-yolo,” Sensors, vol. 24, no. 19, 2024. [Online]. Available: https://www.mdpi.com/1424-8220/24/19/6268

    work page 2024

  2. [41]

    Steel surface defect detection based on multi-layer fusion networks,

    H. Li, M. Liu, Y . Yin, and W. Sun, “Steel surface defect detection based on multi-layer fusion networks,” vol. 15, no. 1, p. 10371, 2025. [Online]. Available: https://www.nature.com/articles/s41598-024-74601-3

    work page 2025

  3. [1]

    Long-term talkingface generation via motion-prior conditional diffusion model,

    F. Shen, C. Wang, J. Gao, Q. Guo, J. Dang, J. Tang, and T.-S. Chua, “Long-term talkingface generation via motion-prior conditional diffusion model,”arXiv preprint arXiv:2502.09533, 2025

    arXiv 2025

  4. [2]

    Imagharmony: Controllable image editing with consistent object quantity and layout,

    F. Shen, X. Du, Y . Gao, J. Yu, Y . Cao, X. Lei, and J. Tang, “Imagharmony: Controllable image editing with consistent object quantity and layout,”arXiv preprint arXiv:2506.01949, 2025

    arXiv 2025

  5. [3]

    Rich feature hierarchies for accurate object detection and semantic segmentation,

    R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 580–587

    2014

  6. [4]

    Fast r-cnn,

    R. Girshick, “Fast r-cnn,” in 2015 IEEE International Conference on Computer Vision (ICCV), 2015, pp. 1440–1448

    work page 2015

  7. [5]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    R. Shaoqing, H. Kaiming, G. Ross, and S. Jian, “Faster r-cnn: Towards real-time object detection with region proposal networks,” Proc. Adv. Neural Inf. Process. Syst., pp. 1–26, 2015

    work page 2015

  8. [6]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r-cnn,” in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2961–2969

    2017

Show all 48 references
  1. [7]

    Cascade r-cnn: Delving into high quality object detection,

    Z. Cai and N. Vasconcelos, “Cascade r-cnn: Delving into high quality object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 6154–6162

    2018

  2. [8]

    Region proposal by guided anchoring,

    J. Wang, K. Chen, S. Yang, C. C. Loy, and D. Lin, “Region proposal by guided anchoring,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 2965–2974

    2019

  3. [9]

    Cascade rpn: Delving into high-quality region proposal network with adaptive convolution,

    T. Vu, H. Jang, T. X. Pham, and C. Yoo, “Cascade rpn: Delving into high-quality region proposal network with adaptive convolution,” Advances in neural information processing systems, vol. 32, 2019

    2019

  4. [10]

    Ssd: Single shot multibox detector,

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y . Fu, and A. C. Berg, “Ssd: Single shot multibox detector,” inComputer Vision–ECCV 2016: 14th European Conference, Am- sterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14. Springer, 2016, pp. 21–37

    2016

  5. [11]

    Yolov3: An incremental improvement,

    J. Redmon and A. Farhadi, “Yolov3: An incremental improvement,” arXiv preprint arXiv:1804.02767, 2018

    2018 arXiv

  6. [12]

    Yolov4: Optimal speed and accuracy of object detection,

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

    2004 arXiv

  7. [13]

    Yolov1 to v8: Unveiling each variant–a comprehensive review of yolo,

    M. Hussain, “Yolov1 to v8: Unveiling each variant–a comprehensive review of yolo,”IEEE Access, vol. 12, pp. 42 816–42 833, 2024

    2024

  8. [14]

    Metal surface defect detection using SLF-YOLO enhanced YOLOv8 model,

    Y . Liu, Y . Liu, X. Guo, X. Ling, and Q. Geng, “Metal surface defect detection using SLF-YOLO enhanced YOLOv8 model,” vol. 15, no. 1, p. 11105, 2025. [Online]. Available: https://www.nature.com/articles/s41598-025-94936-9

    2025

  9. [15]

    Aff-net: A strip steel surface defect detec- tion network via adaptive focusing features,

    Y . Du, H. Chen, Y . Fu, J. Zhu, and H. Zeng, “Aff-net: A strip steel surface defect detec- tion network via adaptive focusing features,” IEEE Transactions on Instrumentation and Measurement, 2024

    2024

  10. [16]

    Multi-scale ship target detection using sar images based on improved yolov5,

    M. Yasir, L. Shanwei, X. Mingming, S. Hui, M. S. Hossain, A. T. I. Colak, D. Wang, W. Jianhua, and K. B. Dang, “Multi-scale ship target detection using sar images based on improved yolov5,” Frontiers in Marine Science, vol. 9, p. 1086140, 2023

    2023

  11. [17]

    Yolo-lfpd: A lightweight method for strip surface defect detection,

    J. Lu, M. Zhu, K. Qin, and X. Ma, “Yolo-lfpd: A lightweight method for strip surface defect detection,” Biomimetics, vol. 9, no. 10, p. 607, 2024. Title Suppressed Due to Excessive Length 13

    2024

  12. [18]

    Imaggarment-1: Fine-grained gar- ment generation for controllable fashion design,

    F. Shen, J. Yu, C. Wang, X. Jiang, X. Du, and J. Tang, “Imaggarment-1: Fine-grained gar- ment generation for controllable fashion design,” arXiv preprint arXiv:2504.13176, 2025

    2025 arXiv

  13. [19]

    Imagdressing-v1: Customizable virtual dressing,

    F. Shen, X. Jiang, X. He, H. Ye, C. Wang, X. Du, Z. Li, and J. Tang, “Imagdressing-v1: Customizable virtual dressing,” in Proceedings of the AAAI Conference on Artificial Intel- ligence, vol. 39, no. 7, 2025, pp. 6795–6804

    2025

  14. [20]

    Imagpose: A unified conditional framework for pose-guided person generation,

    F. Shen and J. Tang, “Imagpose: A unified conditional framework for pose-guided person generation,” Advances in neural information processing systems , vol. 37, pp. 6246–6266, 2024

    2024

  15. [21]

    You only look once: Unified, real-time object detection,

    J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779–788

    2016

  16. [22]

    YOLOv5., in : https://github.com/ultralytics/yolov5

  17. [23]

    Yolo-world: Real-time open- vocabulary object detection,

    T. Cheng, L. Song, Y . Ge, W. Liu, X. Wang, and Y . Shan, “Yolo-world: Real-time open- vocabulary object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 16 901–16 911

    2024

  18. [24]

    QCF-YOLO: A lightweight model of surface defect detection for quick-connect fittings,

    L. Zhou, S. Yang, C. Wang, P. Huang, S. Wang, Y . Wang, and Q. Wang, “QCF-YOLO: A lightweight model of surface defect detection for quick-connect fittings,” vol. 25, no. 1, pp. 1716–1731, 2025. [Online]. Available: https://ieeexplore.ieee.org/document/10752899/

    2025

  19. [25]

    A novel cross frequency-domain interaction learning for aerial oriented object detection,

    W. Weng, W. Lin, F. Lin, J. Ren, and F. Shen, “A novel cross frequency-domain interaction learning for aerial oriented object detection,” inChinese Conference on Pattern Recognition and Computer Vision (PRCV). Springer, 2023, pp. 292–305

    2023

  20. [26]

    A novel multi-frequency coordinated module for sar ship detection,

    C. Qiao, F. Shen, X. Wang, R. Wang, F. Cao, S. Zhao, and C. Li, “A novel multi-frequency coordinated module for sar ship detection,” in2022 IEEE 34th International Conference on Tools with Artificial Intelligence (ICTAI). IEEE, 2022, pp. 804–811

    2022

  21. [27]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing sys- tems, vol. 30, 2017

    2017

  22. [28]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132–7141

    2018

  23. [29]

    Cbam: Convolutional block attention module,

    S. Woo, J. Park, J.-Y . Lee, and I. S. Kweon, “Cbam: Convolutional block attention module,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 3–19

    2018

  24. [30]

    Yolo-hmc: An improved method for pcb surface defect detection,

    M. Yuan, Y . Zhou, X. Ren, H. Zhi, J. Zhang, and H. Chen, “Yolo-hmc: An improved method for pcb surface defect detection,”IEEE Transactions on Instrumentation and Measurement, vol. 73, pp. 1–11, 2024

    2024

  25. [31]

    Enhancing aerial object detection with selective frequency interaction network,

    W. Weng, M. Wei, J. Ren, and F. Shen, “Enhancing aerial object detection with selective frequency interaction network,”IEEE Transactions on Artificial Intelligence, vol. 1, no. 01, pp. 1–12, 2024

    2024

  26. [32]

    Lr-fpn: Enhancing remote sensing object detection with location refined feature pyramid network,

    H. Li, R. Zhang, Y . Pan, J. Ren, and F. Shen, “Lr-fpn: Enhancing remote sensing object detection with location refined feature pyramid network,”arXiv preprint arXiv:2404.01614, 2024

    2024 arXiv

  27. [33]

    dataset., in : https://github.com/lvxiaoming2019/GC10-DET-metallic-surface-defect- datasets

    G.-D. dataset., in : https://github.com/lvxiaoming2019/GC10-DET-metallic-surface-defect- datasets

  28. [34]

    Weakly supervised learning of a classifier for unusual event detection,

    M. Jager, C. Knoll, and F. A. Hamprecht, “Weakly supervised learning of a classifier for unusual event detection,”IEEE Transactions on Image Processing, vol. 17, no. 9, pp. 1700– 1708, 2008

    2008

  29. [35]

    DAGM2007, in : https://hci.iwr.uni-heidelberg.de/content/weakly-supervised-learning- industrial-optical-inspection, 2023

    2023

  30. [36]

    Joining spatial deformable con- volution and a dense feature pyramid for surface defect detection,

    Z. Huang, C. Zhang, L. Ge, Z. Chen, K. Lu, and C. Wu, “Joining spatial deformable con- volution and a dense feature pyramid for surface defect detection,” IEEE Transactions on Instrumentation and Measurement, 2024. 14 J. Hu

    2024

  31. [37]

    Es-net: Efficient scale-aware network for tiny defect detection,

    X. Yu, W. Lyu, D. Zhou, C. Wang, and W. Xu, “Es-net: Efficient scale-aware network for tiny defect detection,” IEEE Transactions on Instrumentation and Measurement , vol. 71, pp. 1–14, 2022

    2022

  32. [38]

    Cspnet: A new backbone that can enhance learning capability of cnn,

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

    2020

  33. [39]

    Spatial pyramid pooling in deep convolutional net- works for visual recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Spatial pyramid pooling in deep convolutional net- works for visual recognition,” IEEE transactions on pattern analysis and machine intelli- gence, vol. 37, no. 9, pp. 1904–1916, 2015

    1904

  34. [42]

    Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,

    C.-Y . Wang, A. Bochkovskiy, and H.-Y . M. Liao, “Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,” in Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, 2023, pp. 7464–7475

    2023

  35. [43]

    Object detection method for grasping robot based on improved yolov5,

    Q. Song, S. Li, Q. Bai, J. Yang, X. Zhang, Z. Li, and Z. Duan, “Object detection method for grasping robot based on improved yolov5,” Micromachines, vol. 12, no. 11, p. 1273, 2021

    2021

  36. [44]

    Steel surface defect detection based on mobilevitv2 and yolov8,

    Z. Lv, Z. Zhao, K. Xia, G. Gu, K. Liu, and X. Chen, “Steel surface defect detection based on mobilevitv2 and yolov8,” The Journal of Supercomputing, vol. 80, no. 13, pp. 18 919– 18 941, 2024

    2024

  37. [45]

    Msb r-cnn: A multi-stage balanced defect detection network,

    Z. Xu, S. Lan, Z. Yang, J. Cao, Z. Wu, and Y . Cheng, “Msb r-cnn: A multi-stage balanced defect detection network,” Electronics, vol. 10, no. 16, p. 1924, 2021

    1924

  38. [46]

    Hic-yolov5: Improved yolov5 for small object detection,

    S. Tang, S. Zhang, and Y . Fang, “Hic-yolov5: Improved yolov5 for small object detection,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) , 2024, pp. 6614–6619

    2024

  39. [47]

    Chained cascade network for object detec- tion,

    W. Ouyang, K. Wang, X. Zhu, and X. Wang, “Chained cascade network for object detec- tion,” in Proceedings of the IEEE International Conference on Computer Vision, 2017, pp. 1938–1946

    2017

  40. [48]

    Yolov9: Learning what you want to learn using programmable gradient information,

    C.-Y . Wang, I.-H. Yeh, and H.-Y . Mark Liao, “Yolov9: Learning what you want to learn using programmable gradient information,” in European conference on computer vision . Springer, 2024, pp. 1–21

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.