Pith. sign in

REVIEW 2 major objections 8 minor 70 references

Cascaded Multi-Scale Attention for Enhanced Multi-Scale Feature Extraction and Interaction with Low-Resolution Images

T0 review · 2 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Cascaded multi-scale attention lets small CNN-ViT hybrids beat much larger models on tiny, low-resolution images.

desk verdict A genuinely new attention combo with strong low-resolution gains that survives its evaluation caveats; send it to review with a request for matched-schedule comparisons and robustness tests. read the letter →

arxiv 2412.02197 v4 pith:NBQ4WV44 submitted 2024-12-03 cs.CV

Xiangyong Lu , Masanori Suganuma , Takayuki Okatani This is my paper · ORCID
classification cs.CV
keywords low-resolutionimagerecognitioncascadedmulti-scaleattentionwindow-basedself-attentionhumanposeestimationheadclassificationCNN-ViThybridfeaturefusion
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 argues that the standard recipe for multi-scale features—downsampling feature maps to several resolutions—is self-defeating when the input image is already tiny, because each downsampling step discards information that cannot be recovered. It proposes a cascaded multi-scale attention (CMSA) mechanism for CNN-ViT hybrids that creates multi-scale features at a single resolution: groups of attention heads attend over windows of different sizes, and the output of each group is fed into the next group so scales interact. On 32x24 human-pose crops, the largest CMSA model reports 56.4 AP on COCO val, above ViTPose-B's 51.2 AP while using about 12 times fewer parameters; on 32x32 head-pose images it reports 4.46 MAE on AFLW2000, better than several methods that run at 224x224. If these results hold, the mechanism offers a parameter-efficient inductive bias for recognition from small images, with applications to surveillance and edge-device inference.

What carries the argument

The central mechanism is cascaded multi-scale attention (CMSA), a drop-in replacement for the multi-head self-attention block in a CNN-ViT hybrid. CMSA divides the attention heads into groups; the first group runs self-attention over the whole feature map, and subsequent groups run window-based local self-attention with progressively smaller windows, so each group operates at a different scale without any downsampling. The output of each group is concatenated with the next group's keys and values and passed through a channel-fusion pointwise convolution and a spatial-fusion depthwise convolution, so information flows from coarser to finer scales and is mixed into the attention computation; all group outputs are then concatenated and linearly projected into the block output. The mechanism is what lets the model cover global and local context at the same feature resolution, avoiding the downsampling the paper identifies as harmful for low-resolution inputs.

What would settle it

Train CMSA and the strongest baseline (for example, ViTPose-B for pose) on COCO crops degraded by realistic low-resolution simulation—blur, sensor noise, and downsampling—instead of clean resizing, and compare at 32x24; if the baseline closes the reported 5.2 AP gap, the claimed advantage of cascaded multi-scale attention for true low-resolution imagery is not established.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a CNN-ViT hybrid can extract and fuse multi-scale features without ever downsampling the feature map inside an attention stage: different attention heads are assigned to different spatial scales, and outputs from coarser-scale groups are passed into finer-scale groups before attention is computed. CMSA splits queries, keys, and values into head groups, applies full-image attention in the first group and Swin-style window attention with progressively smaller windows in later groups, and updates each group's keys and values by concatenating the previous group's output and passing it through channel-fusion and spatial-fusion convolutions. The authors report consistent gains over state-of-the-art pose and classification models at low resolutions, with the largest variant using 7.3M parameters and reaching 56.4 AP on COCO val at 32x24 input. The claim is that grouped cascaded window attention is an effective inductive bias specifically for low-resolution inputs, where conventional downsampling-based multi-scale designs lose critical detail.

Load-bearing premise

The paper's comparisons rest on the assumption that resizing clean full-resolution images down to 32x24 or 32x32 faithfully represents the low-resolution imaging conditions it targets, so the conclusions may not carry over if real low-resolution captures add blur, noise, or aliasing that these experiments do not model.

Editorial extensions

If this is right

  • At 32x24 COCO val, CMSA-L reaches 56.4 AP against 51.2 for ViTPose-B while using roughly 12 times fewer parameters, so the reported advantage grows as input resolution drops.
  • On 32x32 head-pose input, CMSA-L achieves 4.46 MAE on AFLW2000, beating several 224x224 methods, including TokenHPE-v2 at 4.66 MAE.
  • On CIFAR-100 at 32x32, CMSA-L reaches 85.2% top-1 accuracy with 7.4M parameters, above SBCFormer-L's 84.1% with 18.2M parameters.
  • Ablation results attribute the gains to all four components: standard attention gives 48.3 AP on 32x24 COCO val, grouped attention 51.8, adding the cascade 53.5, spatial fusion 54.4, and the full CMSA 56.4.

Reading between the lines

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

  • If the core claim is right, the same grouped, cascade-fed window attention could transfer to other dense prediction tasks on small inputs—such as low-resolution semantic segmentation or object detection—where multi-scale reasoning is needed but feature-map downsampling is costly.
  • The paper evaluates by resizing clean high-resolution crops to tiny sizes; a natural test is to add realistic low-resolution degradation (blur, noise, compression) at train and test time, where window-based local attention may interact with degradation differently than clean downsampling.
  • Because CMSA avoids global attention and keeps feature maps at one resolution, its FLOP and speed numbers suggest it could be attractive for edge-device inference; the paper motivates that use case but does not measure on-device latency, which would be a direct follow-up.
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

2 major / 8 minor

Summary. The paper proposes Cascaded Multi-Scale Attention (CMSA), an attention mechanism for CNN-ViT hybrid networks operating on low-resolution images. CMSA groups the heads of multi-head self-attention, assigns a different local window size to each group, and cascades each group's output into the next group's key/value fusion through channel and spatial fusion modules, thereby enabling multi-scale feature interaction without intra-stage downsampling. The authors instantiate three model sizes (CMSA-S/B/L) and evaluate them on COCO 2017 human pose estimation at 128x96, 64x48, and 32x24; on head-pose estimation using BIWI and AFLW2000 at 64x64 and 32x32; and on CIFAR-10/100 at 32x32. They report consistent improvements over prior state-of-the-art methods with substantially fewer parameters, e.g., 56.4 AP at 32x24 for CMSA-L versus 51.2 AP for ViTPose-B, and 85.2% CIFAR-100 accuracy for CMSA-L versus 84.1% for SBCFormer-L. An ablation study (Table V) shows monotonic AP improvements from 48.3 to 56.4 as each component is added. Code is provided.

Significance. If the results hold, the paper demonstrates a useful inductive bias for low-resolution vision: grouped window attention with cascaded multi-scale feature fusion outperforms standard multi-scale backbones at tiny input resolutions while using an order of magnitude fewer parameters. The ablation supports the mechanism, the results span three tasks, and the paper is open about its design debts to SG-Former and EfficientViT. It also provides code, FLOPs, and measured speeds, which are strengths. The main caveat is that the headline COCO comparison at 32x24 is not scheduler-matched; the 64x48 matched-schedule rows show a small AP cost, so the size of the 32x24 margin under a controlled protocol is currently unquantified. With a matched-schedule row added at 32x24, the central claim would be solid.

major comments (2)
  1. [Table II / §IV-A-3] The 32x24 column is the central evidence for the low-resolution claim, but the CMSA rows in that column use a different learning-rate schedule than the baselines; the matched-step-scheduler rows (CMSA-B* and CMSA-L*) are reported only at 64x48. At 64x48, switching to the baseline scheduler changes CMSA-B AP from 65.2 to 64.7 and CMSA-L AP from 66.0 to 65.9, while AR moves in the opposite direction. Therefore the headline 5.2-point AP margin over ViTPose-B at 32x24 could be partly a scheduler artifact. Please add matched-schedule runs at 32x24 and state in the table or caption which scheduler each row uses.
  2. [§IV-A-2 / §IV-A-3] The baseline numbers are produced by running official repositories at resolutions those papers did not test, with no indication of whether any hyperparameters were re-tuned for the new resolution. Because the paper's central claim is quantitative (outperforming state-of-the-art methods with fewer parameters), the training recipe for every baseline row should be reported, including the scheduler, number of epochs, learning rate, augmentation, and whether any hyperparameter search was performed. A fully matched-recipe comparison at 32x24 for all methods would make the reported margins directly interpretable.
minor comments (8)
  1. [Abstract / §III-A] The abstract and Section III-A state that CMSA avoids downsampling 'of the input image or feature maps,' but the architecture uses patch-embedding stages that halve the resolution (Fig. 2a, §III-D-1). Footnote 1 qualifies this to within-stage processing; please make the abstract wording consistent with that qualification.
  2. [§III-C, Eq. (1)] The text says the lower stream updates 'K_k and Q_k,' but Eq. (1) concatenates K_k, V_k, and X'_{k-1} and updates K and V only; Q_k is not updated. Please correct the text or the equation.
  3. [§III-D-2] The text says 'For the FNN' but the block uses an FFN (feed-forward network); please fix the typo.
  4. [§IV-A-2] AdamW does not have a momentum hyperparameter in the usual sense; if the intended setting is beta1 = 0.9, please state that explicitly.
  5. [§IV-C-2] The sentence 'It shows averaged values over 300 trials' is confusing; presumably the models were trained for 300 epochs and the table reports single-run accuracies. Please clarify and, if multiple runs were averaged, report the variance.
  6. [§IV-B-3 / Table III] The statement that CMSA-L at 32x32 achieves an MAE of 4.46 on AFLW2000, 'surpassing the performance of most typical models for higher image resolutions,' is not supported by the best higher-resolution rows in the same table (e.g., 6DRepNet at 224x224 gives 3.61 and Li et al. at 224x224 gives 3.99). Please restrict the claim to the low-resolution comparison or specify the subset of models to which it refers.
  7. [§III-C, optional setting] When the spatial-fusion module halves the spatial size of K and V, the relationship between the query window size (s_k x t_k) and the reduced key/value grid is not specified. Please add a sentence describing how local attention is computed in this optional configuration.
  8. [Table II, footnote 6] The footnote 'they tend to show slightly lower AR scores, it is attributable to the choice of learning rate schedulers' is grammatically unclear; it should be rewritten and should state explicitly which rows in the table use which scheduler.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CMSA's empirical claims are direct benchmark measurements, with only non-load-bearing self-citations.

full rationale

The paper makes no formal derivation whose output is equivalent to its input by construction; its claims are direct empirical measurements on public benchmarks (COCO 2017, BIWI/AFLW2000, CIFAR-10/100). The CMSA mechanism is explicitly defined by the equations and block diagrams in Section III, and the reported accuracies are obtained by training and evaluating the resulting networks, not by substituting the target outcome into the construction. The design is admittedly inspired by external works (SG-Former, EfficientViT, Swin Transformer), and the authors' own SBCFormer appears only as a related-work reference and as a comparison baseline in Figures 3-4 and Table IV; no load-bearing premise is justified solely by self-citation. The ablation in Table V independently supports the contribution of each architectural component. The strongest caveat is the unmatched training recipe for baselines and the scheduler effect disclosed in footnote 6 of Section IV-A-3; that is a comparison-protocol concern, not a circularity, because the CMSA numbers are measured rather than forced by construction. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is repackaged under new coordinates. Therefore no circular step can be identified.

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

Architecture hyperparameters (window sizes, group counts, and the optional KV reduction) are chosen by hand and are the main free inputs; they are not derived from first principles. The paper's domain assumptions are that multi-scale features help low-resolution recognition and that intra-stage downsampling is harmful. No invented entities are introduced.

free parameters (3)
  • Per-group window sizes (s_k, t_k) = Stage 1: (32,24), (16,12), (8,6); Stage 2/3: (16,12), (8,6) or (8,6), (8,6) (Table I)
    Hand-chosen scale levels for each attention group; the paper does not provide a principle for selecting these exact sizes, so they are free design inputs.
  • Group architecture (number of groups, channel dims, heads per group) = 3 groups (stage 1), 2 groups (stages 2 and 3); dims 16 to 64, heads 1 to 5 (Table I)
    Capacity allocation is hand-designed per model size; no derivation from a computational budget is given.
  • Optional 2x2 average pooling on K and V in SF = Enabled in experiments
    The optional setting halves key/value spatial size and doubles value channels; its effects on accuracy are not separately ablated, so it is a hand-selected efficiency choice.
assumptions (3)
  • domain assumption Multi-scale feature extraction and interaction improve accuracy for pose estimation and image recognition.
    Presented as established motivation in Section III-A with citations [1]-[3]; the paper assumes this transfers to low-resolution inputs.
  • domain assumption Intra-stage downsampling of feature maps is harmful for low-resolution images, so a no-downsampling attention design is beneficial.
    Central motivation in Sections I and III-A; footnote 1 restricts the claim to within-stage processing, so the axiom is specifically about attention-level multi-scale handling.
  • domain assumption The DeiT training recipe (210/300 epochs, AdamW, augmentation) is appropriate for all compared models at low resolutions.
    Used for CMSA and, with official repos, for baselines (Section IV-A-2); if baselines require different schedules at low resolution, comparisons could be biased.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cascaded Multi-Scale Attention for Enhanced Multi-Scale Feature Extraction and Interaction with Low-Resolution Images." pith.science (2026). https://pith.science/paper/NBQ4WV44

@misc{pith2026241202197,
  author       = {Pith},
  title        = {Pith review of: Cascaded Multi-Scale Attention for Enhanced Multi-Scale Feature Extraction and Interaction with Low-Resolution Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NBQ4WV44}},
  note         = {Machine review of arXiv:2412.02197}
}
read the original abstract

In real-world applications of image recognition tasks, such as human pose estimation, cameras often capture objects, like human bodies, at low resolutions. This scenario poses a challenge in extracting and leveraging multi-scale features, which is often essential for precise inference. To address this challenge, we propose a new attention mechanism, named cascaded multi-scale attention (CMSA), tailored for use in CNN-ViT hybrid architectures, to handle low-resolution inputs effectively. The design of CMSA enables the extraction and seamless integration of features across various scales without necessitating the downsampling of the input image or feature maps. This is achieved through a novel combination of grouped multi-head self-attention mechanisms with window-based local attention and cascaded fusion of multi-scale features over different scales. This architecture allows for the effective handling of features across different scales, enhancing the model's ability to perform tasks such as human pose estimation, head pose estimation, and more with low-resolution images. Our experimental results show that the proposed method outperforms existing state-of-the-art methods in these areas with fewer parameters, showcasing its potential for broad application in real-world scenarios where capturing high-resolution images is not feasible. Code is available at https://github.com/xyongLu/CMSA.

Figures

Figures reproduced from arXiv: 2412.02197 by the authors.

Figure 1
Figure 1. Left: Estimating pose from low-resolution images. Left-Upper: human [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Design of the proposed model. (a) The overall architecture. (b) Design of the basic blocks. (c) Proposed cascaded multi-scale attention (CMSA). See [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Average Precision vs. Parameters. All models are trained for human [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Accuracy vs. Parameters. All models are trained for CIFAR-10 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Heatmap visualizations of the left wrist, elbow, and knee from models [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Attention map visualizations from CMSA modules at each stage of [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 46 canonical work pages

  1. [1]

    High- erhrnet: Scale-aware representation learning for bottom-up human pose estimation,

    B. Cheng, B. Xiao, J. Wang, H. Shi, T. S. Huang, and L. Zhang, “High- erhrnet: Scale-aware representation learning for bottom-up human pose estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 5386–5395

    2020

  2. [2]

    Hrformer: High-resolution vision transformer for dense predict,

    Y . Yuan, R. Fu, L. Huang, W. Lin, C. Zhang, X. Chen, and J. Wang, “Hrformer: High-resolution vision transformer for dense predict,” Ad- vances in Neural Information Processing Systems , vol. 34, pp. 7281– 7293, 2021

    work page 2021

  3. [3]

    Lite pose: Efficient architecture design for 2d human pose estimation,

    Y . Wang, M. Li, H. Cai, W.-M. Chen, and S. Han, “Lite pose: Efficient architecture design for 2d human pose estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 13 126–13 136

    work page 2022

  4. [4]

    Tempose: A new skeleton-based transformer model designed for fine-grained motion recognition in badminton,

    M. Ibh, S. Grasshof, D. Witzner, and P. Madeleine, “Tempose: A new skeleton-based transformer model designed for fine-grained motion recognition in badminton,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 5198–5207

    work page 2023

  5. [5]

    Hiformer: Hierarchical multi-scale representations using transformers for medical image segmentation,

    M. Heidari, A. Kazerouni, M. Soltany, R. Azad, E. K. Aghdam, J. Cohen-Adad, and D. Merhof, “Hiformer: Hierarchical multi-scale representations using transformers for medical image segmentation,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023, pp. 6202–6212

    2023

  6. [6]

    Deep high-resolution repre- sentation learning for human pose estimation,

    K. Sun, B. Xiao, D. Liu, and J. Wang, “Deep high-resolution repre- sentation learning for human pose estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 5693–5703

    2019

  7. [7]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

    2021

  8. [8]

    Sg-former: Self-guided transformer with evolving token reallocation,

    S. Ren, X. Yang, S. Liu, and X. Wang, “Sg-former: Self-guided transformer with evolving token reallocation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 6003–6014

    work page 2023

Show all 70 references
  1. [9]

    Efficientvit: Memory efficient vision transformer with cascaded group attention,

    X. Liu, H. Peng, N. Zheng, Y . Yang, H. Hu, and Y . Yuan, “Efficientvit: Memory efficient vision transformer with cascaded group attention,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 14 420–14 430

    2023

  2. [10]

    Crossvit: Cross-attention multi- scale vision transformer for image classification,

    C.-F. R. Chen, Q. Fan, and R. Panda, “Crossvit: Cross-attention multi- scale vision transformer for image classification,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 357– 366

    2021

  3. [11]

    Few-shot object detection with fully cross-transformer,

    G. Han, J. Ma, S. Huang, L. Chen, and S.-F. Chang, “Few-shot object detection with fully cross-transformer,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 5321– 5330

    2022

  4. [12]

    Topformer: Token pyramid transformer for mobile semantic segmentation,

    W. Zhang, Z. Huang, G. Luo, T. Chen, X. Wang, W. Liu, G. Yu, and C. Shen, “Topformer: Token pyramid transformer for mobile semantic segmentation,” in Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , 2022, pp. 12 083–12 093

    2022

  5. [13]

    P2t: Pyramid pooling transformer for scene understanding,

    Y .-H. Wu, Y . Liu, X. Zhan, and M.-M. Cheng, “P2t: Pyramid pooling transformer for scene understanding,” IEEE transactions on pattern analysis and machine intelligence , vol. 45, no. 11, pp. 12 760–12 771, 2022

    2022

  6. [14]

    Shunted self-attention via multi-scale token aggregation,

    S. Ren, D. Zhou, S. He, J. Feng, and X. Wang, “Shunted self-attention via multi-scale token aggregation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 10 853–10 862

    2022

  7. [15]

    Pyramid vision transformer: A versatile backbone for dense prediction without convolutions,

    W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pyramid vision transformer: A versatile backbone for dense prediction without convolutions,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 568–578

    2021

  8. [16]

    Scale-aware modula- tion meet transformer,

    W. Lin, Z. Wu, J. Chen, J. Huang, and L. Jin, “Scale-aware modula- tion meet transformer,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 6015–6026

    2023

  9. [17]

    Sbcformer: Lightweight network capable of full-size imagenet classification at 1 fps on single board computers,

    X. Lu, M. Suganuma, and T. Okatani, “Sbcformer: Lightweight network capable of full-size imagenet classification at 1 fps on single board computers,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 1123–1133

    2024

  10. [18]

    Task-driven super resolu- tion: Object detection in low-resolution images,

    M. Haris, G. Shakhnarovich, and N. Ukita, “Task-driven super resolu- tion: Object detection in low-resolution images,” in Neural Information Processing: 28th International Conference, ICONIP 2021, Sanur, Bali, Indonesia, December 8–12, 2021, Proceedings, Part V 28 . Springer, ...

    2021

  11. [19]

    Derivenet for (very) low resolution image classification,

    M. Singh, S. Nagpal, R. Singh, and M. Vatsa, “Derivenet for (very) low resolution image classification,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 10, pp. 6569–6577, 2021

    2021

  12. [20]

    Image super-resolution with non-local sparse attention,

    Y . Mei, Y . Fan, and Y . Zhou, “Image super-resolution with non-local sparse attention,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 3517– 3526

    2021

  13. [21]

    Low-resolution human pose estimation,

    C. Wang, F. Zhang, X. Zhu, and S. S. Ge, “Low-resolution human pose estimation,” Pattern Recognition, vol. 126, p. 108579, 2022

    2022

  14. [22]

    Feature map distillation of thin nets for low-resolution object recognition,

    Z. Huang, S. Yang, M. Zhou, Z. Li, Z. Gong, and Y . Chen, “Feature map distillation of thin nets for low-resolution object recognition,” IEEE Transactions on Image Processing , vol. 31, pp. 1364–1379, 2022

    2022

  15. [23]

    Transformer for single image super-resolution,

    Z. Lu, J. Li, H. Liu, C. Huang, L. Zhang, and T. Zeng, “Transformer for single image super-resolution,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 457– 466

    2022

  16. [24]

    Essaformer: Efficient transformer for hyperspectral image super- resolution,

    M. Zhang, C. Zhang, Q. Zhang, J. Guo, X. Gao, and J. Zhang, “Essaformer: Efficient transformer for hyperspectral image super- resolution,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 23 073–23 084

    2023

  17. [25]

    Steformer: Efficient stereo image super- resolution with transformer,

    J. Lin, L. Yin, and Y . Wang, “Steformer: Efficient stereo image super- resolution with transformer,” IEEE Transactions on Multimedia, vol. 25, pp. 8396–8407, 2023

    2023

  18. [26]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” in 9th International Conference on Le...

    2021

  19. [27]

    Cswin transformer: A general vision transformer backbone with cross-shaped windows,

    X. Dong, J. Bao, D. Chen, W. Zhang, N. Yu, L. Yuan, D. Chen, and B. Guo, “Cswin transformer: A general vision transformer backbone with cross-shaped windows,” in Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition , 2022, pp. 12 124– 12 134

    2022

  20. [28]

    Neighborhood attention transformer,

    A. Hassani, S. Walton, J. Li, S. Li, and H. Shi, “Neighborhood attention transformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 6185–6194

    2023

  21. [29]

    Vision transformer with quad- rangle attention,

    Q. Zhang, J. Zhang, Y . Xu, and D. Tao, “Vision transformer with quad- rangle attention,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  22. [30]

    Mobilevit: Light-weight, general-purpose, and mobile-friendly vision transformer,

    S. Mehta and M. Rastegari, “Mobilevit: Light-weight, general-purpose, and mobile-friendly vision transformer,” in The Tenth International Conference on Learning Representations , 2022

    2022

  23. [31]

    Rethinking vision transformers for mobilenet size and speed,

    Y . Li, J. Hu, Y . Wen, G. Evangelidis, K. Salahi, Y . Wang, S. Tulyakov, and J. Ren, “Rethinking vision transformers for mobilenet size and speed,” arXiv preprint arXiv:2212.08059 , 2022

    2022 arXiv

  24. [32]

    Rethinking mobile block for efficient neural models,

    J. Zhang, X. Li, J. Li, L. Liu, Z. Xue, B. Zhang, Z. Jiang, T. Huang, Y . Wang, and C. Wang, “Rethinking mobile block for efficient neural models,” arXiv preprint arXiv:2301.01146 , 2023

    2023 arXiv

  25. [33]

    Fastvit: A fast hybrid vision transformer using structural reparameterization,

    P. K. A. Vasu, J. Gabriel, J. Zhu, O. Tuzel, and A. Ranjan, “Fastvit: A fast hybrid vision transformer using structural reparameterization,” arXiv preprint arXiv:2303.14189, 2023

    2023 arXiv

  26. [34]

    Repvit: Revisiting mobile cnn from vit perspective,

    A. Wang, H. Chen, Z. Lin, J. Han, and G. Ding, “Repvit: Revisiting mobile cnn from vit perspective,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 15 909–15 920

    2024

  27. [35]

    Hydra atten- tion: Efficient attention with many heads,

    D. Bolya, C.-Y . Fu, X. Dai, P. Zhang, and J. Hoffman, “Hydra atten- tion: Efficient attention with many heads,” in European Conference on Computer Vision. Springer, 2022, pp. 35–49

    2022

  28. [36]

    The principle of diversity: Training stronger vision transformers calls for reducing all levels of redundancy,

    T. Chen, Z. Zhang, Y . Cheng, A. Awadallah, and Z. Wang, “The principle of diversity: Training stronger vision transformers calls for reducing all levels of redundancy,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 12 020–12 030

    2022

  29. [37]

    Levit: a vision transformer in convnet’s clothing for faster inference,

    B. Graham, A. El-Nouby, H. Touvron, P. Stock, A. Joulin, H. J ´egou, and M. Douze, “Levit: a vision transformer in convnet’s clothing for faster inference,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 12 259–12 269. IEEE TRANSACTIONS O...

    2021

  30. [38]

    Pvt v2: Improved baselines with pyramid vision transformer,

    W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pvt v2: Improved baselines with pyramid vision transformer,” Computational Visual Media , vol. 8, no. 3, pp. 415–424, 2022

    2022

  31. [39]

    Cmt: Convolutional neural networks meet vision transformers,

    J. Guo, K. Han, H. Wu, Y . Tang, X. Chen, Y . Wang, and C. Xu, “Cmt: Convolutional neural networks meet vision transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 12 175–12 185

    2022

  32. [40]

    Con- ditional positional encodings for vision transformers,

    X. Chu, Z. Tian, B. Zhang, X. Wang, X. Wei, H. Xia, and C. Shen, “Con- ditional positional encodings for vision transformers,” arXiv preprint arXiv:2102.10882, 2021

    2021 arXiv

  33. [41]

    Diverse branch block: Building a convolution as an inception-like unit,

    X. Ding, X. Zhang, J. Han, and G. Ding, “Diverse branch block: Building a convolution as an inception-like unit,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 10 886–10 895

    2021

  34. [42]

    Mobileone: An improved one millisecond mobile backbone,

    P. K. A. Vasu, J. Gabriel, J. Zhu, O. Tuzel, and A. Ranjan, “Mobileone: An improved one millisecond mobile backbone,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 7907–7917

    2023

  35. [43]

    Simple baselines for human pose estimation and tracking,

    B. Xiao, H. Wu, and Y . Wei, “Simple baselines for human pose estimation and tracking,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 466–481

    2018

  36. [44]

    Distribution-aware coordinate representation for human pose estimation,

    F. Zhang, X. Zhu, H. Dai, M. Ye, and C. Zhu, “Distribution-aware coordinate representation for human pose estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 7093–7102

    2020

  37. [45]

    The devil is in the details: Delving into unbiased data processing for human pose estimation,

    J. Huang, Z. Zhu, F. Guo, and G. Huang, “The devil is in the details: Delving into unbiased data processing for human pose estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 5700–5709

    2020

  38. [46]

    Vitpose: Simple vision transformer baselines for human pose estimation,

    Y . Xu, J. Zhang, Q. Zhang, and D. Tao, “Vitpose: Simple vision transformer baselines for human pose estimation,” Advances in Neural Information Processing Systems , vol. 35, pp. 38 571–38 584, 2022

    2022

  39. [47]

    Human pose as compositional tokens,

    Z. Geng, C. Wang, Y . Wei, Z. Liu, H. Li, and H. Hu, “Human pose as compositional tokens,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 660–671

    2023

  40. [48]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 . Springer,...

    2014

  41. [49]

    Training data-efficient image transformers & distillation through attention,

    H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablayrolles, and H. J ´egou, “Training data-efficient image transformers & distillation through attention,” in International Conference on Machine Learning . PMLR, 2021, pp. 10 347–10 357

    2021

  42. [50]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in 7th International Conference on Learning Representations, ICLR , 2019

    2019

  43. [51]

    Face alignment across large poses: A 3d solution,

    X. Zhu, Z. Lei, X. Liu, H. Shi, and S. Z. Li, “Face alignment across large poses: A 3d solution,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 146–155

    2016

  44. [52]

    Random forests for real time 3d face analysis,

    G. Fanelli, M. Dantone, J. Gall, A. Fossati, and L. Van Gool, “Random forests for real time 3d face analysis,” International journal of computer vision, vol. 101, pp. 437–458, 2013

    2013

  45. [53]

    Fine-grained head pose estimation without keypoints,

    N. Ruiz, E. Chong, and J. M. Rehg, “Fine-grained head pose estimation without keypoints,” in Proceedings of the IEEE conference on computer vision and pattern recognition workshops , 2018, pp. 2074–2083

    2018

  46. [54]

    Fsa-net: Learning fine-grained structure aggregation for head pose estimation from a single image,

    T.-Y . Yang, Y .-T. Chen, Y .-Y . Lin, and Y .-Y . Chuang, “Fsa-net: Learning fine-grained structure aggregation for head pose estimation from a single image,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 1087–1096

    2019

  47. [55]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and 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

  48. [56]

    Improving head pose estimation with a combined loss and bounding box margin adjustment,

    M. Shao, Z. Sun, M. Ozay, and T. Okatani, “Improving head pose estimation with a combined loss and bounding box margin adjustment,” in 2019 14th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2019) . IEEE, 2019, pp. 1–5

    2019

  49. [57]

    Whenet: Real-time fine-grained estimation for wide range head pose,

    Y . Zhou and J. Gregson, “Whenet: Real-time fine-grained estimation for wide range head pose,” in 31st British Machine Vision Conference 2020, BMVC 2020, Virtual Event, UK, September 7-10, 2020 , 2020

    2020

  50. [58]

    Tokenhpe: Learning orientation tokens for efficient head pose estimation via transformers,

    C. Zhang, H. Liu, Y . Deng, B. Xie, and Y . Li, “Tokenhpe: Learning orientation tokens for efficient head pose estimation via transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 8897–8906

    2023

  51. [59]

    Fdn: Feature decoupling net- work for head pose estimation,

    H. Zhang, M. Wang, Y . Liu, and Y . Yuan, “Fdn: Feature decoupling net- work for head pose estimation,” in Proceedings of the AAAI conference on artificial intelligence , vol. 34, no. 07, 2020, pp. 12 789–12 796

    2020

  52. [60]

    Accurate head pose estima- tion using image rectification and a lightweight convolutional neural network,

    X. Li, D. Zhang, M. Li, and D.-J. Lee, “Accurate head pose estima- tion using image rectification and a lightweight convolutional neural network,” IEEE Transactions on Multimedia , 2022

    2022

  53. [61]

    Eva-gcn: Head pose estimation based on graph convolutional networks,

    M. Xin, S. Mo, and Y . Lin, “Eva-gcn: Head pose estimation based on graph convolutional networks,” in Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition, 2021, pp. 1462– 1471

    2021

  54. [62]

    Mfdnet: Collaborative poses perception and matrix fisher distribution for head pose estimation,

    H. Liu, S. Fang, Z. Zhang, D. Li, K. Lin, and J. Wang, “Mfdnet: Collaborative poses perception and matrix fisher distribution for head pose estimation,” IEEE Transactions on Multimedia , vol. 24, pp. 2449– 2460, 2021

    2021

  55. [63]

    Real-time 6dof full-range markerless head pose estimation,

    R. Algabri, H. Shin, and S. Lee, “Real-time 6dof full-range markerless head pose estimation,” Expert Systems with Applications , vol. 239, p. 122293, 2024

    2024

  56. [64]

    Toward robust and unconstrained full range of rotation head pose estimation,

    T. Hempel, A. A. Abdelrahman, and A. Al-Hamadi, “Toward robust and unconstrained full range of rotation head pose estimation,” IEEE Transactions on Image Processing , vol. 33, pp. 2377–2387, 2024

    2024

  57. [65]

    A vector-based representation to enhance head pose estimation,

    Z. Cao, Z. Chu, D. Liu, and Y . Chen, “A vector-based representation to enhance head pose estimation,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2021, pp. 1188–1197

    2021

  58. [66]

    2dheadpose: A simple and effective annotation method for the head pose in rgb images and its dataset,

    Y . Wang, W. Zhou, and J. Zhou, “2dheadpose: A simple and effective annotation method for the head pose in rgb images and its dataset,” Neural Networks, vol. 160, pp. 50–62, 2023

    2023

  59. [67]

    Tokens-to-token vit: Training vision transformers from scratch on imagenet,

    L. Yuan, Y . Chen, T. Wang, W. Yu, Y . Shi, Z.-H. Jiang, F. E. Tay, J. Feng, and S. Yan, “Tokens-to-token vit: Training vision transformers from scratch on imagenet,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 558–567

    2021

  60. [68]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 11 976–11 986

    2022

  61. [69]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778. Dr . Xiangyong Lu received the Ph.D. degree in the Computer Vision Lab, Graduate School of...

    2016

  62. [2017]

    His research interests are in the field of computer vision and machine learning

    He was an Assistant Professor at Tohoku University and currently is a research scientist at Sakana AI in Japan. His research interests are in the field of computer vision and machine learning. Dr . Takayuki Okatani earned his B.Eng., M.Sc., and Ph.D. degrees in Mathematical En...

    1994

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.