Pith. sign in

REVIEW 5 major objections 4 minor 37 references

Local-Global Attention: An Adaptive Mechanism for Multi-Scale Feature Integration

T0 review · 5 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Local-Global Attention fuses fine-grained local features from three convolution scales with a larger-kernel global branch plus positional encoding, using learned alpha weights, and the paper reports consistent mAP gains over standard…

desk verdict Plausible module, unproven claims: the experimental section fails to establish consistent gains, with contradicted baselines and missing details. read the letter →

arxiv 2411.09604 v1 pith:NPEMA4O2 submitted 2024-11-14 cs.CV cs.AI

Yifan Shao This is my paper · ORCID
classification cs.CVcs.AI
keywords objectdetectionattentionmechanismlocal-globalmulti-scalefeatureintegrationpositionalencodingsmallYOLOv8MobileNetV3
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 introduces Local-Global Attention, a plug-in module that runs a local attention branch and a global attention branch in parallel and then fuses them with learnable scalar weights. The local branch uses multi-scale convolutions with kernel sizes 3, 5, and 7 plus residual connections; the global branch uses a larger convolution kernel with positional encoding. The learned $\alpha_\text{local}$ and $\alpha_\text{global}$ weights let the network shift the balance between detail and context for each input. The author reports that, across nine object detection and classification benchmarks and three backbones, the module beats existing attention mechanisms such as SE, CBAM, ECA, MHSA, and SA while keeping FLOPs and parameters at roughly baseline levels. A sympathetic reader should care because the claim is that the long-standing local/global trade-off in attention can be handled cheaply and adaptively, with the clearest gains on small-object and multi-class datasets.

What carries the argument

The load-bearing object is the two-branch attention head with a learned fusion. The local branch applies depthwise and $1\times1$ convolutions at kernel sizes $k \in \{3,5,7\}$ to a position-encoded, scale-weighted feature map, computes attention scores $Q_k K_k^\top / \sqrt{d}$, averages the three branch outputs, and adds a residual. The global branch applies the same attention formula with one larger kernel $k_g$ to the same input. The outputs are combined as $\text{out} = \alpha_\text{local}\,\text{local\_out} + \alpha_\text{global}\,\text{global\_out}$, where the alphas are learned, and a final $1\times1$ convolution compresses the fusion. This mechanism is what carries the argument: the residuals preserve identity features, the multiple kernel sizes supply granularity, the positional encoding supplies spatial structure, and the alphas make the local/global balance data-dependent without adding a learned projection or extra FLOPs.

What would settle it

Inspect the released code for the global kernel size and positional encoding; then run the TinyPerson ablation with the global branch removed and with alpha fixed at 0.5 instead of learned, and if mAP does not drop or change, the global branch and the adaptive fusion are not the source of the reported gains.

Watch

Extended reading notes

Core claim

The paper's central claim is that combining multi-scale local attention with a positional-encoded global branch, fused by learnable alpha parameters, gives a better balance of fine detail and broad context than any single attention mechanism. On the paper's own terms this is established empirically: LGA posts the best or tied-best number on every benchmark, with the largest margins on small-object datasets such as TinyPerson (+0.92 mAP@50 on MobileNetV3, +0.7 on YOLOv8) and strong results on VOC2012 (+0.7 mAP@50-95), VisDrone2019, DOTAv1.0, and COCOminitrain, while GFLOPs stay equal to the baseline in nearly every row. The adaptive alpha fusion is the specific reason offered for these gains: the model can emphasize local detail when the input needs it and global context when it needs that.

Load-bearing premise

The paper assumes the larger-kernel branch actually captures distinct global context, but it never specifies the global kernel size or the positional encoding formula, so the two branches could be encoding overlapping information and the learned alpha weights could be adjusting noise rather than a real local/global trade-off.

Editorial extensions

If this is right

  • Plugging LGA into MobileNetV3, ResNet18, or YOLOv8 improves detection mAP on small-object benchmarks such as TinyPerson, VisDrone2019, and DOTAv1.0.
  • The module adds essentially no FLOPs and only a few hundred thousand parameters, so it can replace SE, CBAM, or other attention blocks in existing backbones without a compute budget change.
  • The same mechanism improves Top-1 accuracy on MNIST and Fashion-MNIST classification, indicating the local/global fusion is not specific to detection.
  • Because the alpha weights are learned, the module's behavior can adapt per dataset or per input, which is the mechanism behind the claimed multi-scale robustness.

Reading between the lines

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

  • A natural follow-up is repeated-seed evaluation, since the reported margins are 0.1 to 0.9 mAP and the paper does not report variance; such runs would show whether the advantage over CBAM and SE is stable.
  • Visualizing the learned alpha values per image would test the adaptivity claim directly: small-object-heavy inputs should weight the local branch more, and scene-heavy inputs should weight the global branch more.
  • The same two-branch fusion pattern could transfer to semantic segmentation and video object detection, where scale variance and the local/global context trade-off are equally central.
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 Local-Global Attention (LGA), a plug-in attention module that combines multi-scale convolutions (local branch) with a larger-kernel global branch plus positional encoding, and fuses the two branches via learnable alpha parameters. The authors evaluate LGA by inserting it into MobileNetV3, ResNet18, and YOLOv8 backbones for classification (MNIST, Fashion-MNIST), object detection (VOC2012, VisDrone2019, COCOminitrain, DOTAv1.0, GWHD2020), and small-object detection (TinyPerson), comparing against MHSA, SE, CBAM, ECA, and SA. The central claim is that LGA 'consistently outperforms' existing attention mechanisms while maintaining computational efficiency (Abstract and Section 5). The mechanism itself is described with simple equations (Section 3), but several implementation details are unspecified, and the experimental evidence contains internal inconsistencies and lacks statistical support.

Significance. If the claims were substantiated, LGA would be a useful lightweight attention module for detection and classification, with the alpha-fusion idea being a sensible way to balance local and global features. However, the paper's central claim of consistent outperformance is directly contradicted by one of its own tables (SE beats LGA on DOTAv1.0 mAP50), and the reported baselines show signs of serious implementation problems (MHSA degrades VOC2012 by 6.3 mAP, an implausibly large drop). The computational-efficiency claim is unsupported because GFLOPs are identical across very different modules, and the global branch is incompletely specified. The study also reports no error bars, no multiple seeds, and no significance tests, while all improvements are 0.1–0.9 mAP, within typical run-to-run noise. These issues collectively mean the paper's empirical foundation does not establish its headline conclusion. No code is provided despite the statement that it is. The mechanism's design is reasonable, but the validation as presented is not reliable.

major comments (5)
  1. [Table 7 (and Abstract/Section 5)] The claim that LGA 'consistently outperforms' existing attention mechanisms is contradicted by the paper's own results: on DOTAv1.0, SE achieves mAP50=54.4 while LGA achieves 54.3 (Table 7). Since the abstract and Section 5 specifically state consistency, this single counterexample invalidates the headline claim as currently worded.
  2. [Section 4.5.1, Table 6] The MHSA baseline results are implausible and suggest an implementation error. Adding MHSA to MobileNetV3 drops VOC2012 mAP50 from 43.7 to 37.4 (a 6.3-point decrease) and COCOminitrain mAP50 from 14.6 to 11.7 (a 2.9-point decrease). No mainstream attention module, especially a simple MHSA add-on, is known to cause such catastrophic degradation. If the MHSA baseline is incorrectly implemented, then every comparison against MHSA (and by extension the 'outperforms' claim) is unreliable.
  3. [Tables 6 and 7, with Section 3] The GFLOPs values are reported as identical across all methods in nearly every table (e.g., 2.8 for MobileNetV3, MHSA, SE, CBAM, and LGA in Tables 6 and 7), despite LGA adding multiple multi-scale convolution branches, two attention computations, and a 1x1 convolution as described in Section 3. Either the complexity figures were not actually measured (e.g., copied from a single run) or the module's cost is misreported. In either case, the statement that LGA 'maintains computational efficiency' is not supported by the data presented.
  4. [Sections 3.4 and 3.6.2] The global attention branch is not sufficiently specified: the 'larger convolution kernel' size kg is never given, and the positional encoding PE is defined only as 'aligns with the spatial dimensions' without providing the encoding formula or how it is combined with the features. Without these details, the method cannot be reproduced, and the claim that the global branch actually captures global context cannot be verified.
  5. [Section 4 (all experimental tables)] No error bars, multiple random seeds, or statistical significance tests are reported anywhere in the experimental section. The reported improvements range from 0.1 to 0.9 mAP points, which is well within the run-to-run variance typically observed in object detection. The paper's claim of consistent improvement across datasets is therefore not statistically substantiated.
minor comments (4)
  1. [Section 2.3] There is a typo in the heading: 'queeze-and-Excitation Attention' should be 'Squeeze-and-Excitation Attention'.
  2. [Abstract and Section 1] The phrase 'we introduce a learnable parameters' (Abstract) and 'we introduce learnable α parameters' (Section 1) contain grammar errors; the intended meaning is 'learnable parameters'.
  3. [Section 3.1] The tensor notation X ∈ R^{B×D×H×W} is unconventional: D is used for 'embedding dimension' but the standard channel axis is usually C. Also, the text says 'D: Embedding dimension' but then uses 'depth' interchangeably, which is confusing. Please clarify the notation.
  4. [Abstract and Section 1] The paper states 'Code is available at the link' but provides no actual URL. Since the method is claimed to be reproducible, a working link or repository identifier should be given.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Local-Global Attention mechanism is evaluated on external benchmarks, and its learnable parameters are trained by standard supervised learning rather than fitted to the reported metric.

full rationale

The paper's derivation chain defines an attention mechanism in Equations (1)-(15), then reports empirical results on external datasets (MNIST, Fashion-MNIST, TinyPerson, VOC2012, VisDrone2019, COCOminitrain, DOTAv1.0, GWHD2020). The learnable alpha parameters in Equations (5), (6), and (14) are trained on data and are not constructed so that the reported mAP scores follow by definition. No target metric is defined in terms of the model's own outputs, and no fitted parameter is renamed as a prediction. The paper does not rely on self-citations: all cited works are standard external references, and the method is not justified by an author-imported uniqueness theorem. The weaknesses noted by the reader and skeptic--unreliable baseline behavior in Tables 6 and 7, identical GFLOPs across different modules, missing code link, unspecified global kernel size and positional encoding formula, and the fact that SE outperforms LGA on DOTAv1.0 mAP50--are experimental-reporting and validity concerns, not circularity. They may undermine the strength of the empirical claims, but they do not make the derivation equivalent to its inputs. Consistent with the default expectation for papers evaluated against external benchmarks, the circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. The free parameters are mostly unspecified architectural hyperparameters (kernel sizes, number of scales, heads) and learned fusion weights. The key domain assumptions are that YOLOv8 defaults are fair across all datasets and that the alpha-weighted fusion of local and global branches is effective; neither is justified beyond the weak empirical results.

free parameters (5)
  • local kernel sizes = {3, 5, 7}
    Hand-chosen in Eq. (7); no ablation or justification for these specific sizes.
  • global kernel size
    Referenced as 'larger convolution kernel' in Eq. (12), but size never specified; the global context claim depends on this unspecified choice.
  • number of scales
    Eq. (4) defines a set of scales but never states how many; affects architecture but unspecified.
  • number of attention heads
    Eq. (1) divides D by num heads; num heads is never given.
  • alpha_local and alpha_global = learned values not reported
    Eq. (14) uses learned scalar weights for fusion; no initialization or normalization constraint is provided.
assumptions (4)
  • domain assumption YOLOv8 default training settings are appropriate for all datasets
    Stated in Sec. 4.2, the paper uses YOLOv8 defaults for all datasets without evidence that they are optimal for each; unfair defaults for baselines would undermine comparisons.
  • domain assumption The benchmark datasets are representative of small-object detection difficulty
    The paper claims strong performance on small object detection (TinyPerson, VisDrone, DOTA) but provides no scale analysis to connect the method's design to the claimed multi-scale benefit.
  • standard math Depthwise convolutions with residual connections extract useful local features
    Standard CNN building block in Eq. (3); assumed valid without proof.
  • ad hoc to paper The learned alpha-weighted sum of local and global outputs is an effective fusion strategy
    Eq. (14) is the core of the method, but no theoretical or empirical comparison against other fusion methods (concatenation, gating, etc.) is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Local-Global Attention: An Adaptive Mechanism for Multi-Scale Feature Integration." pith.science (2026). https://pith.science/paper/NPEMA4O2

@misc{pith2026241109604,
  author       = {Pith},
  title        = {Pith review of: Local-Global Attention: An Adaptive Mechanism for Multi-Scale Feature Integration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NPEMA4O2}},
  note         = {Machine review of arXiv:2411.09604}
}
read the original abstract

In recent years, attention mechanisms have significantly enhanced the performance of object detection by focusing on key feature information. However, prevalent methods still encounter difficulties in effectively balancing local and global features. This imbalance hampers their ability to capture both fine-grained details and broader contextual information-two critical elements for achieving accurate object detection.To address these challenges, we propose a novel attention mechanism, termed Local-Global Attention, which is designed to better integrate both local and global contextual features. Specifically, our approach combines multi-scale convolutions with positional encoding, enabling the model to focus on local details while concurrently considering the broader global context. Additionally, we introduce a learnable parameters, which allow the model to dynamically adjust the relative importance of local and global attention, depending on the specific requirements of the task, thereby optimizing feature representations across multiple scales.We have thoroughly evaluated the Local-Global Attention mechanism on several widely used object detection and classification datasets. Our experimental results demonstrate that this approach significantly enhances the detection of objects at various scales, with particularly strong performance on multi-class and small object detection tasks. In comparison to existing attention mechanisms, Local-Global Attention consistently outperforms them across several key metrics, all while maintaining computational efficiency.

Figures

Figures reproduced from arXiv: 2411.09604 by the authors.

Figure 1
Figure 1. To validate the effectiveness of our method, we conducted extensive experiments on multiple bench￾mark datasets, including VOC2007 [6], VOC2012 [7], VisDrone2019-DET [37], TinyPerson [34], COCO2017 [16], GWHD2020 [3], COCO minitrain [23], DOTA-v1.0 [30], as well as MNIST [4] and Fashion-MNIST [31]. The results demonstrate that Local-Global Attention outper￾forms existing attention mechanisms with similar computa￾tio… view at source ↗
Figure 2
Figure 2. Local-global attention structure diagram [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 23 canonical work pages

  1. [1]

    Neural machine translation by jointly learning to align and translate

    Dzmitry Bahdanau. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473, 2014. 1

    arXiv 2014

  2. [2]

    Yolov4: Optimal speed and accuracy of object detection, 2020

    Alexey Bochkovskiy, Chien-Yao Wang, and Hong- Yuan Mark Liao. Yolov4: Optimal speed and accuracy of object detection, 2020. 2

    work page 2020

  3. [3]

    Global Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methods

    Etienne David, Simon Madec, Pouria Sadeghi-Tehran, Helge Aasen, Bangyou Zheng, Shouyang Liu, Norbert Kirchgess- ner, Goro Ishikawa, Koichi Nagasawa, Minhajul Badhon, et al. Global wheat head detection (gwhd) dataset: A large and diverse dataset of high-resolution rgb-labelled images to develop and benchmark wheat head detection methods. arXiv preprint arX...

    work page Pith review arXiv 2005

  4. [4]

    The mnist database of handwritten digit images for machine learning research

    Li Deng. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Maga- zine, 29(6):141–142, 2012. 2, 6

    work page 2012

  5. [5]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 3

    arXiv 2010

  6. [6]

    Everingham, L

    M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman. The PASCAL Visual Object Classes Challenge 2007 (VOC2007) Results. http://www.pascal- network.org/challenges/VOC/voc2007/workshop/index.html. 2

    work page 2007

  7. [7]

    Everingham, L

    M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman. The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results. http://www.pascal- network.org/challenges/VOC/voc2012/workshop/index.html. 2, 6

    work page 2012

  8. [8]

    Fast r-cnn

    R Girshick. Fast r-cnn. arXiv preprint arXiv:1504.08083 ,

    arXiv

Show all 37 references
  1. [9]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 2, 6, 7

    2016

  2. [10]

    Searching for mo- bilenetv3

    Andrew Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, et al. Searching for mo- bilenetv3. In Proceedings of the IEEE/CVF international conference on computer vision, pages 1314–1324, 2019. 2, 6, 7

    2019

  3. [11]

    Mobilenets: Efficient convolutional neu- ral networks for mobile vision applications

    Andrew G Howard. Mobilenets: Efficient convolutional neu- ral networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017. 2

    2017 arXiv

  4. [12]

    Squeeze-and-excitation net- works

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net- works. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 7132–7141, 2018. 1, 2, 3

    2018

  5. [13]

    Ultralytics yolov5, 2020

    Glenn Jocher. Ultralytics yolov5, 2020. 2

    2020

  6. [14]

    Ultralytics yolov8, 2023

    Glenn Jocher, Ayush Chaurasia, and Jing Qiu. Ultralytics yolov8, 2023. 2, 6, 7

    2023

  7. [15]

    Yolov6 v3.0: A full-scale reloading, 2023

    Chuyi Li, Lulu Li, Yifei Geng, Hongliang Jiang, Meng Cheng, Bo Zhang, Zaidan Ke, Xiaoming Xu, and Xiangx- iang Chu. Yolov6 v3.0: A full-scale reloading, 2023. 2

    2023

  8. [16]

    Lawrence Zitnick, and Piotr Doll ´ar

    Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Doll ´ar. Microsoft coco: Common objects in context, 2015. 2, 5

    2015

  9. [17]

    Ssd: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. Ssd: Single shot multibox detector. In Computer Vision–ECCV 2016: 14th European Conference, Amster- dam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14,...

    2016

  10. [18]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 1

    2021

  11. [19]

    Effective approaches to attention- based neural machine translation

    Minh-Thang Luong. Effective approaches to attention- based neural machine translation. arXiv preprint arXiv:1508.04025, 2015. 1

    2015 arXiv

  12. [20]

    You only look once: Unified, real-time object de- tection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object de- tection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016. 1

    2016

  13. [21]

    Yolov3: An incremental improvement

    Joseph Redmon and Ali Farhadi. Yolov3: An incremental improvement. arXiv preprint arXiv:1804.02767, 2018. 2

    2018 arXiv

  14. [22]

    An overview of gradient descent optimiza- tion algorithms

    Sebastian Ruder. An overview of gradient descent optimiza- tion algorithms. arXiv preprint arXiv:1609.04747, 2016. 7

    2016 arXiv

  15. [23]

    Hough- net: Integrating near and long-range evidence for bottom-up object detection

    Nermin Samet, Samet Hicsonmez, and Emre Akbas. Hough- net: Integrating near and long-range evidence for bottom-up object detection. In European Conference on Computer Vi- sion (ECCV), 2020. 2, 6

    2020

  16. [24]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zh- moginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 4510–4520, 2018. 2

    2018

  17. [25]

    Attention is all you need

    Ashish Vaswani. Attention is all you need. arXiv preprint arXiv:1706.03762, 2017. 1, 3

    2017 arXiv

  18. [26]

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

    Chien-Yao Wang, Alexey Bochkovskiy, and Hong- Yuan Mark Liao. Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. arXiv preprint arXiv:2207.02696, 2022. 2

    2022 arXiv

  19. [27]

    Eca-net: Efficient channel at- tention for deep convolutional neural networks

    Qilong Wang, Banggu Wu, Pengfei Zhu, Peihua Li, Wang- meng Zuo, and Qinghua Hu. Eca-net: Efficient channel at- tention for deep convolutional neural networks. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11534–11542, 2020. 1

    2020

  20. [28]

    Cbam: Convolutional block attention module

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block attention module. In Proceedings of the European conference on computer vision (ECCV), pages 3–19, 2018. 1, 3

    2018

  21. [29]

    Fbnet: Hardware-aware efficient con- vnet design via differentiable neural architecture search

    Bichen Wu, Xiaoliang Dai, Peizhao Zhang, Yanghan Wang, Fei Sun, Yiming Wu, Yuandong Tian, Peter Vajda, Yangqing Jia, and Kurt Keutzer. Fbnet: Hardware-aware efficient con- vnet design via differentiable neural architecture search. In Proceedings of the IEEE/CVF conference on c...

    2019

  22. [30]

    Dota: A large-scale dataset for object detection in 9 aerial images

    Gui-Song Xia, Xiang Bai, Jian Ding, Zhen Zhu, Serge Be- longie, Jiebo Luo, Mihai Datcu, Marcello Pelillo, and Liang- pei Zhang. Dota: A large-scale dataset for object detection in 9 aerial images. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2...

    2018

  23. [31]

    Fashion- mnist: a novel image dataset for benchmarking machine learning algorithms, 2017

    Han Xiao, Kashif Rasul, and Roland V ollgraf. Fashion- mnist: a novel image dataset for benchmarking machine learning algorithms, 2017. 2, 6

    2017

  24. [32]

    Focal self-attention for local-global interactions in vision transformers

    Jianwei Yang, Chunyuan Li, Pengchuan Zhang, Xiyang Dai, Bin Xiao, Lu Yuan, and Jianfeng Gao. Focal self-attention for local-global interactions in vision transformers. arXiv preprint arXiv:2107.00641, 2021. 1, 3

    2021 arXiv

  25. [33]

    Hierarchical attention networks for document classification

    Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. Hierarchical attention networks for document classification. In Proceedings of the 2016 con- ference of the North American chapter of the association for computational linguistics: human language tec...

    2016

  26. [34]

    Scale match for tiny person detection

    Xuehui Yu, Yuqi Gong, Nan Jiang, Qixiang Ye, and Zhenjun Han. Scale match for tiny person detection. In Proceedings of the IEEE/CVF winter conference on applications of com- puter vision, pages 1257–1265, 2020. 2, 6, 7

    2020

  27. [35]

    Manmatha, Mu Li, and Alexander Smola

    Hang Zhang, Chongruo Wu, Zhongyue Zhang, Yi Zhu, Haibin Lin, Zhi Zhang, Yue Sun, Tong He, Jonas Mueller, R. Manmatha, Mu Li, and Alexander Smola. Resnest: Split- attention networks. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) W...

    2022

  28. [36]

    Sa-net: Shuffle atten- tion for deep convolutional neural networks

    Qing-Long Zhang and Yu-Bin Yang. Sa-net: Shuffle atten- tion for deep convolutional neural networks. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 2235–2239. IEEE, 2021. 1

    2021

  29. [37]

    Detection and tracking meet drones challenge

    Pengfei Zhu, Longyin Wen, Dawei Du, Xiao Bian, Heng Fan, Qinghua Hu, and Haibin Ling. Detection and tracking meet drones challenge. IEEE Transactions on Pattern Anal- ysis and Machine Intelligence, 44(11):7380–7399, 2021. 2, 6 10

    2021

Pith tools

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