Pith. sign in

REVIEW 5 major objections 5 minor 24 references

SWinMamba: Serpentine Window State Space Model for Vascular Segmentation

T0 review · 5 major / 5 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 A serpentine window ordering through a bidirectional state space model keeps slender vessels connected, beating ten baselines on three medical datasets while using only 2.78M parameters and 8.20 GFLOPs.

desk verdict Real architecture, unproven numeric claim: the serpentine-window Mamba idea is worth a look, but the reported gains are point estimates without variance or a stated loss on tiny datasets. read the letter →

arxiv 2507.01323 v1 pith:5K2XE6O2 submitted 2025-07-02 eess.IV cs.CVcs.LG

Rongchang Zhao , Huanchi Liu , Jian Zhang This is my paper
classification eess.IVcs.CVcs.LG
keywords vascularsegmentationstatespacemodelMambaserpentinewindowtokenizerbidirectionalaggregationspatial-frequencyfusionmedicalimage
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

Vascular segmentation in medical images fails when slender vessels are split into disconnected pieces, and this paper claims that the cause is how the image is turned into tokens. The proposed SWinMamba network orders overlapping local windows into serpentine sequences that follow the vessel direction, then feeds them through a bidirectional state space model so the segmenter sees a vessel as one continuous tubular object. On three datasets --- retinal fundus, OCT angiography, and X-ray coronary angiography --- the authors report that SWinMamba beats ten published methods on most metrics, including the largest average gain of $3.15\%$ on $\beta_0$, the connectivity error, while using only $2.78$M parameters and $8.20$ GFLOPs. The paper's contribution, read sympathetically, is a tokenization-and-scanning recipe that replaces rigid raster scans with anatomy-aligned sequences, making state space models usable for continuity-critical segmentation.

What carries the argument

The load-bearing machinery is the Serpentine Window Tokenizer (SWToken) combined with the Bidirectional Aggregation Module (BAM). SWToken computes strings of anchor points whose lateral offsets are predicted by a convolution, so the $s\times s$ windows overlap and bend along the vessel direction instead of sitting on a fixed grid; fractional sampling is carried out by bilinear interpolation. BAM arranges the tokens of each window into a sequence, runs a forward and a backward state space scan across windows, and reassembles the processed patches into a feature map. A third component, the Spatial-Frequency Fusion Unit (SFFU), runs the same windowed processing in the Fourier domain and fuses it with the spatial branch under attention weights. Together, these convert slender, meandering vessels into ordered local contexts that a state space model can reason about sequentially.

What would settle it

Retrain SWinMamba and the strongest baseline from each family with at least ten random seeds under the identical augmentation, optimizer, and schedule, and compare the distributions of $\beta_0$, clDice, and Dice; the claimed $3.15\%$, $0.39\%$, and $0.14\%$ improvements would be falsified if the gaps fall within one standard deviation of the paired differences.

Watch

Extended reading notes

Core claim

The paper's central claim is that continuity, not just pixel accuracy, is what a vessel segmenter should be structured around, and that a state space model can deliver it if the scanning path follows the vessel. SWinMamba is said to produce complete and connected vascular segmentation by replacing the rigid 1D flattening of Vision Mamba with overlapping serpentine window sequences: anchor points are shifted by learnable offsets so local windows snake along vessel contours, and a bidirectional aggregation module scans those windows forward and backward to integrate coherent local features. A Spatial-Frequency Fusion Unit then fuses spatial-domain and frequency-domain features under attention weights. With $2.78$M parameters and $8.20$ GFLOPs, the reported results on CHASE-DB1, OCTA-500, and DCA1 beat all ten compared methods on most metrics, with average improvements of $3.15\%$ in $\beta_0$, $0.39\%$ in clDice, and $0.14\%$ in Dice.

Load-bearing premise

The load-bearing premise is that the reported gains over ten baselines are larger than run-to-run noise, since the evidence is four-fold cross-validation on small datasets with single point estimates, no variance or significance tests, and no stated training loss.

Editorial extensions

If this is right

  • If SWinMamba is right, connectivity errors in vessel segmentation can be reduced by aligning sequence order with anatomy rather than by increasing model size.
  • The network outputs connected vessels directly, so downstream analysis of diameter, curvature, and bifurcation angles would not need a separate repair step.
  • At $2.78$M parameters and $8.20$ GFLOPs, the approach is light enough for clinical workstations, unlike transformer baselines that require hundreds of GFLOPs.
  • Since the largest reported gain is on $\beta_0$, the design targets the topological error that matters most for quantifying vascular geometry.

Reading between the lines

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

  • A natural extension is 3D: the same anchor-string construction could snake through CT or MR angiography volumes, where vessels are curves in three dimensions and connectivity errors are even harder to measure.
  • The frequency-domain branch suggests that phase information carries continuity cues; one testable ablation would remove only the phase of the Fourier features and watch what happens to the $\beta_0$ gain.
  • Because the training loss and seed variance are not reported, the most direct test of the contribution is to retrain the best baseline and SWinMamba under several seeds with the identical loss and schedule, turning the claimed averages into distributions.
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 / 5 minor

Summary. The paper proposes SWinMamba, a U-shaped encoder-decoder for vascular segmentation with three main components: a Serpentine Window Tokenizer (SWToken) that constructs learnable overlapping window strings along vascular structures, a Bidirectional Aggregation Module (BAM) that feeds these token sequences into a bidirectional state space model, and a Spatial-Frequency Fusion Unit (SFFU) that combines spatial- and frequency-domain features. The method is evaluated on CHASE-DB1, OCTA-500, and DCA1 against ten published baselines using clDice, Betti number error β0, Dice, FLOPs, and parameter counts. The central claim is that SWinMamba outperforms state-of-the-art methods on most metrics with low computational cost, achieving average improvements of 3.15% in β0, 0.39% in clDice, and 0.14% in Dice across the three datasets (Section 3.5, Table 1).

Significance. The architectural idea is timely and the evaluation scope is a strength: three datasets of different modalities, ten baselines spanning convolutional, hybrid, and Mamba-based families, and the stated use of official codes. If the reported superiority is robust, SWinMamba would be a useful low-cost SSM-based alternative for tubular structure segmentation. However, the significance is currently conditional because the main evidence consists of single-run point estimates: the reported Dice margins are far smaller than the expected fold-to-fold variability on a dataset as small as CHASE-DB1, and Table 1 contains cells in which SWinMamba is not the best despite the caption claiming the best results. The submission does not provide code, trained models, or per-fold results, so the empirical claim is not independently verifiable as presented.

major comments (5)
  1. [Section 3.5, Table 1] The central superiority claim is not supported by the reported statistics. Four-fold cross-validation on CHASE-DB1 (28 images) leaves roughly 7 test images per fold; with a per-image Dice standard deviation of even 3 points, the per-fold standard error is above 1 point, which is roughly an order of magnitude larger than the headline 0.14% average Dice gain. Report repeated-seed results as mean±std, confidence intervals, and paired significance tests across folds, and provide per-fold breakdowns for all methods. The same limitation applies to the ablation study in Table 2, where component contributions are single-point estimates and the word 'significant' is used informally.
  2. [Section 3.3, Section 3.5, Table 1] Table 1 is internally inconsistent with its caption stating that SWinMamba achieves the best results. On CHASE-DB1, the β0 of SWinMamba (1.546, lower is better) is worse than DSCNet's 1.434, and on DCA1 the Dice of SWinMamba (0.7659) is worse than DE-DCGCN-EE's 0.7660. Since β0 is the connectivity metric that the paper's premise targets, the aggregate '3.15% improvement in β0' conceals a per-dataset regression on CHASE-DB1; the average is largely carried by OCTA-500, where SWinMamba's 2.636 is well ahead of the next best 3.042. The bolding and the claim of consistent connected-vessel improvement should be corrected.
  3. [Section 3.2] The training loss is never stated. Because clDice is both an evaluation metric and a commonly used training objective for tubular-structure segmentation, omitting the loss makes it impossible to determine whether the reported gains come from the architecture or from a different training objective. State the loss function, any weighting or auxiliary losses, and, if possible, retrain all baselines with the same objective for the main comparison.
  4. [Section 3.2, Section 3.5] The comparison protocol is underspecified. The text says all baselines are implemented using official codes, but it does not state whether they were retrained under the same four-fold split, augmentation, random-crop size, optimizer schedule, and epoch count, or whether Table 1 reuses published numbers. It also does not describe the test-time inference procedure, which is especially important because training uses 256×256 random crops while input images are as large as 999×960. These choices can easily account for differences of the magnitude reported here, so the full protocol must be documented and per-fold results or code should be supplied.
  5. [Section 2.1] The core tokenizer depends on three hand-set parameters (L=9, s=8, α=2), described only as 'set empirically.' No sensitivity analysis is provided, so it is unclear whether SWinMamba's performance is stable in the parameter region or tuned specifically to these datasets. Add an ablation over L, s, and α, or provide a robustness argument.
minor comments (5)
  1. [Section 3.2] Define precisely how the Betti number error β0 is computed, including whether it is a per-image mean, whether it is normalized, and how connected components are counted. This metric is central to the connectivity claim.
  2. [Section 2.1, Eq. (2)] Clarify the notation in Eq. (2): specify the range and construction of the predicted offset field Δy_k, and explain how coordinates that fall outside the feature map are handled during bilinear sampling.
  3. [Section 3.5] Explain how the aggregate 'average improvements' are computed across datasets and metrics, especially for β0 where lower is better and the numerical scale differs between datasets.
  4. [General] The submission does not provide code, trained models, or an explicit reproducibility statement. For an empirical point-estimate comparison, releasing these resources would substantially strengthen the paper.
  5. [Figure 1] Figure 1 contains several short labels and fragments (for example, the text boxes under the challenge illustrations) that are difficult to read in the submitted version; the figure should be redrawn for legibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SWinMamba's central claim is an empirical benchmark against public datasets, with no self-citation chain or fitted-input prediction.

full rationale

The paper's load-bearing claim is that SWinMamba outperforms ten published baselines on CHASE-DB1, OCTA-500, and DCA1 (Table 1, Section 3.5). This is a measurement against external, public datasets with all baselines implemented from their official codes; it does not follow by construction from the method equations. The method's components (SWToken, BAM, SFFU) are defined in Sections 2.1-2.3 as concrete operations (bilinear interpolation, bidirectional SSM scanning, FFT/IFFT with attention), and none of the reported metrics is a fitted parameter renamed as a prediction. The chosen hyperparameters L=9, s=8, alpha=2 are disclosed as empirical settings (Section 2.1), not derived from the target results by definition. References [22] and [24] are external works (CBAM and Vim), and the authors do not cite their own prior results as load-bearing evidence. The known statistical weakness of the evaluation — four-fold single-run point estimates without variance or significance tests — is an evidentiary limitation, not circularity, because the comparison remains externally falsifiable. Therefore no circular step meeting the quoted-evidence standard was found.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The central claim depends on benchmark conditions rather than on a mathematical derivation. The paper supplies no code, no loss function, and no statistical test, so the protocol assumptions carry the result. The only explicit hand-tuned numbers are SWToken constants and training hyperparameters.

free parameters (8)
  • String length L = 9
    Set empirically in Section 2.1; controls how many windows form one serpentine sequence.
  • Window stride s = 8
    Set empirically in Section 2.1; controls window spacing and token set density.
  • Extend stride alpha = 2
    Set empirically in Section 2.1; step size between anchor points.
  • Learning rate = 1e-4
    Adam setting in Section 3.2; no sensitivity analysis.
  • Epochs = 800
    Training schedule in Section 3.2; no convergence criterion given.
  • Batch size = 1
    Chosen due to hardware limits in Section 3.2.
  • Random crop size = 256x256
    Data preprocessing choice in Section 3.2.
  • Channel counts per repetition = 16/32/64
    Architecture capacity choice given in the Figure 2 caption.
assumptions (5)
  • domain assumption Ground truth labels on CHASE-DB1, OCTA-500, and DCA1 are accurate enough to support metric differences of about 0.001.
    Used in Table 1; label noise alone could exceed some of the reported gaps.
  • domain assumption Four-fold cross-validation with batch size 1 and 800 epochs gives stable estimates on datasets of 28 to 300 images.
    Section 3.2; no repeated runs or error bars are reported.
  • domain assumption The unspecified training loss is fixed and identical across all compared methods.
    Section 3.2 lists optimizer, epochs, batch, and augmentation but never states the loss; fair comparison requires it.
  • domain assumption The compared methods were run with official code in equivalent settings as claimed in Section 3.5.
    Section 3.5 states that all baselines were implemented using official codes, but no verification artifacts are provided.
  • domain assumption The predicted offsets in SWToken train to follow vessel morphology through the segmentation loss alone, without vessel-specific supervision.
    Section 2.1, Eq. 2; there is no explicit topology loss or offset supervision, so the serpentine alignment claim rests on learned behavior.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SWinMamba: Serpentine Window State Space Model for Vascular Segmentation." pith.science (2026). https://pith.science/paper/5K2XE6O2

@misc{pith2026250701323,
  author       = {Pith},
  title        = {Pith review of: SWinMamba: Serpentine Window State Space Model for Vascular Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5K2XE6O2}},
  note         = {Machine review of arXiv:2507.01323}
}
read the original abstract

Vascular segmentation in medical images is crucial for disease diagnosis and surgical navigation. However, the segmented vascular structure is often discontinuous due to its slender nature and inadequate prior modeling. In this paper, we propose a novel Serpentine Window Mamba (SWinMamba) to achieve accurate vascular segmentation. The proposed SWinMamba innovatively models the continuity of slender vascular structures by incorporating serpentine window sequences into bidirectional state space models. The serpentine window sequences enable efficient feature capturing by adaptively guiding global visual context modeling to the vascular structure. Specifically, the Serpentine Window Tokenizer (SWToken) adaptively splits the input image using overlapping serpentine window sequences, enabling flexible receptive fields (RFs) for vascular structure modeling. The Bidirectional Aggregation Module (BAM) integrates coherent local features in the RFs for vascular continuity representation. In addition, dual-domain learning with Spatial-Frequency Fusion Unit (SFFU) is designed to enhance the feature representation of vascular structure. Extensive experiments on three challenging datasets demonstrate that the proposed SWinMamba achieves superior performance with complete and connected vessels.

Figures

Figures reproduced from arXiv: 2507.01323 by the authors.

Figure 1
Figure 1. Challenges. Vascular structures exhibit slender and fragile morphology, posing significant challenges for segmentation algorithms. Motivation. The proposed SWin￾Mamba adaptively guides global visual context modeling to the serpentine window sequences to efficiently model the continuity of slender vascular structures with the bidirectional state space models. variations in diameter and direction along vascular struct… view at source ↗
Figure 2
Figure 2. SWinMamba Overview. The proposed SWinMamba adopts a U-shaped struc￾ture where the encoder includes the novel SWinMamba block and feature fusion unit (SFFU). The SWinMamba block incorporates serpentine window sequences to bidi￾rectional state space models to guide the global context modeling to the continuous vascular structure. The SWinMamba block includes a newly designed tokenizer (SWTo￾ken) for flexible receptive… view at source ↗
Figure 3
Figure 3. Visualization shows that SWinMamba achieves complete and connected vascu￾lar segmentation across three modalities of medical images. ments over M2, proving that SFFU facilitates the construction of comprehensive feature representations for complete vascular segmentation. 3.5 Comparison with State-of-the-art (SOTA) Comparison experiments ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

24 extracted references · 15 canonical work pages

  1. [1]

    Applied Sciences 9(24), 5507 (2019)

    Cervantes-Sanchez, F., Cruz-Aceves, I., Hernandez-Aguirre, A., Hernandez- Gonzalez, M.A., Solorio-Meza, S.E.: Automatic segmentation of coronary arter- ies in x-ray angiograms using multiscale analysis and artificial neural networks. Applied Sciences 9(24), 5507 (2019)

    work page 2019

  2. [2]

    arXiv preprint arXiv:2102.04306 (2021)

    Chen, J., Lu, Y., Yu, Q., Luo, X., Adeli, E., Wang, Y., Lu, L., Yuille, A.L., Zhou, Y.:Transunet:Transformersmakestrongencodersformedicalimagesegmentation. arXiv preprint arXiv:2102.04306 (2021)

    arXiv 2021

  3. [3]

    In: 2022 IEEE International Con- ference on Bioinformatics and Biomedicine (BIBM)

    Chen, X., Hu, T., Guo, S.: Mdanet: Multi-direction attention network for curvilin- ear structure segmentation of biomedical images. In: 2022 IEEE International Con- ference on Bioinformatics and Biomedicine (BIBM). pp. 1475–1480. IEEE (2022)

    work page 2022

  4. [4]

    NeuroImage: Clinical30, 102573 (2021)

    Deshpande, A., Jamilpour, N., Jiang, B., Michel, P., Eskandari, A., Kidwell, C., Wintermark, M., Laksari, K.: Automatic segmentation, feature extraction and comparison of healthy and stroke cerebral vasculature. NeuroImage: Clinical30, 102573 (2021)

    work page 2021

  5. [5]

    biocybernetics and biomedical engineering 34(2), 117–124 (2014)

    Franklin, S.W., Rajan, S.E.: Computerized screening of diabetic retinopathy em- ploying blood vessel segmentation in retinal images. biocybernetics and biomedical engineering 34(2), 117–124 (2014)

    work page 2014

  6. [6]

    IEEE Transactions on Biomedical Engineering59(9), 2538–2548 (2012)

    Fraz, M.M., Remagnino, P., Hoppe, A., Uyyanonvara, B., Rudnicka, A.R., Owen, C.G., Barman, S.A.: An ensemble classification-based approach applied to retinal blood vessel segmentation. IEEE Transactions on Biomedical Engineering59(9), 2538–2548 (2012)

    work page 2012

  7. [7]

    Microvascular Research74(2), 72–84 (2007), therapeutic Applications of Angiogenesis and Anti-angiogenesis

    Fukumura, D., Jain, R.K.: Tumor microvasculature and microenvironment: Tar- gets for anti-angiogenesis and normalization. Microvascular Research74(2), 72–84 (2007), therapeutic Applications of Angiogenesis and Anti-angiogenesis

    work page 2007

  8. [8]

    arXiv preprint arXiv:2312.00752 (2023)

    Gu, A., Dao, T.: Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752 (2023)

    arXiv 2023

Show all 24 references
  1. [9]

    Progress in retinal and eye research91, 101095 (2022)

    Hanssen, H., Streese, L., Vilser, W.: Retinal vessel diameters and function in car- diovascular risk and disease. Progress in retinal and eye research91, 101095 (2022)

    2022

  2. [10]

    Computers in Biology and Medicine 172, 108315 (2024)

    Li, J., Gao, G., Yang, L., Liu, Y.: A retinal vessel segmentation network with multiple-dimension attention and adaptive feature fusion. Computers in Biology and Medicine 172, 108315 (2024)

    2024

  3. [11]

    Medical image analysis93, 103092 (2024)

    Li, M., Huang, K., Xu, Q., Yang, J., Zhang, Y., Ji, Z., Xie, K., Yuan, S., Liu, Q., Chen, Q.: Octa-500: a retinal dataset for optical coherence tomography angiogra- phy study. Medical image analysis93, 103092 (2024)

    2024

  4. [12]

    IEEE Transactions on Medical Imaging41(8), 1975–1989 (2022)

    Li, Y., Zhang, Y., Cui, W., Lei, B., Kuang, X., Zhang, T.: Dual encoder-based dynamic-channel graph convolutional network with edge enhancement for reti- nal vessel segmentation. IEEE Transactions on Medical Imaging41(8), 1975–1989 (2022)

    2022

  5. [13]

    IEEE Transactions on Cyber- netics 53(9), 5826–5839 (2023)

    Li, Y., Zhang, Y., Liu, J.Y., Wang, K., Zhang, K., Zhang, G.S., Liao, X.F., Yang, G.: Global transformer and dual local attention network via deep-shallow hierar- chical feature fusion for retinal vessel segmentation. IEEE Transactions on Cyber- netics 53(9), 5826–5839 (2023)

    2023

  6. [14]

    arXiv preprint arXiv:2401.04722 (2024)

    Ma, J., Li, F., Wang, B.: U-mamba: Enhancing long-range dependency for biomed- ical image segmentation. arXiv preprint arXiv:2401.04722 (2024)

    2024 arXiv

  7. [15]

    Medical image analysis67, 101874 (2021) 10 R.Zhao et al

    Mou, L., Zhao, Y., Fu, H., Liu, Y., Cheng, J., Zheng, Y., Su, P., Yang, J., Chen, L., Frangi, A.F., et al.: Cs2-net: Deep learning segmentation of curvilinear structures in medical imaging. Medical image analysis67, 101874 (2021) 10 R.Zhao et al

    2021

  8. [16]

    In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision

    Qi, Y., He, Y., Qi, X., Zhang, Y., Yang, G.: Dynamic snake convolution based on topological geometric constraints for tubular structure segmentation. In: Proceed- ings of the IEEE/CVF International Conference on Computer Vision. pp. 6070– 6079 (2023)

    2023

  9. [17]

    In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, Oc- tober 5-9, 2015, proceedings, part III 18

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomed- ical image segmentation. In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, Oc- tober 5-9, 2015, proceedings, part III 18....

    2015

  10. [18]

    arXiv preprint arXiv:2402.02491 (2024)

    Ruan, J., Li, J., Xiang, S.: Vm-unet: Vision mamba unet for medical image seg- mentation. arXiv preprint arXiv:2402.02491 (2024)

    2024 arXiv

  11. [19]

    Journal of digital imaging35(2), 226–239 (2022)

    Saitta, S., Sturla, F., Caimi, A., Riva, A., Palumbo, M.C., Nano, G., Votta, E., Corte, A.D., Glauber, M., Chiappino, D., et al.: A deep learning-based and fully automated pipeline for thoracic aorta geometric analysis and planning for endovas- cular repair from computed tomog...

    2022

  12. [20]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Shit, S., Paetzold, J.C., Sekuboyina, A., Ezhov, I., Unger, A., Zhylka, A., Pluim, J.P., Bauer, U., Menze, B.H.: cldice-a novel topology-preserving loss function for tubular structure segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog...

    2021

  13. [21]

    IEEE transactions on medical imaging8(1), 78–88 (1989)

    Sun, Y.: Automated identification of vessel contours in coronary arteriograms by an adaptive tracking algorithm. IEEE transactions on medical imaging8(1), 78–88 (1989)

    1989

  14. [22]

    In: Proceedings of the European conference on computer vision (ECCV)

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

    2018

  15. [23]

    IEEE Transactions on Medical Imaging42(12), 3614–3624 (2023)

    Zhang, L.S., Liu, S.Q., Xie, X.L., Zhou, X.H., Hou, Z.G., Wang, C.N., Qu, X.K., Han, W.Z., Ma, X.Y., Song, M.: A novel spatial position prediction navigation sys- tem makes surgery more accurate. IEEE Transactions on Medical Imaging42(12), 3614–3624 (2023)

    2023

  16. [24]

    In: Proceedings of the International Conference on Machine Learning

    Zhu, L., Liao, B., Zhang, Q., Wang, X., Liu, W., Wang, X.: Vision mamba: Efficient visual representation learning with bidirectional state space model. In: Proceedings of the International Conference on Machine Learning. pp. 62429–62442. PMLR (2024)

    2024

Pith tools

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