Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

FEAT: Full-Dimensional Efficient Attention Transformer for Medical Video Generation

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read FEAT is a diffusion transformer for medical video generation that replaces quadratic self-attention with linear-time spatial, temporal, and channel attention, and it reports matching a state-of-the-art model with a small variant and…

desk verdict A plausible efficiency-focused architecture for medical video generation, but the headline performance claim is not yet backed by robust statistics. read the letter →

arxiv 2506.04956 v1 pith:BXXSNT4R submitted 2025-06-05 cs.CV

Huihan Wang , Zhiwen Yang , Hui Zhang , Dan Zhao , Bingzheng Wei , Yan Xu This is my paper
classification cs.CV
keywords medicalvideogenerationdiffusiontransformerlinear-complexityattentionWKVglobalchannelresidualvalueguidanceendoscopysemi-supervisedlearning
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

FEAT is a diffusion transformer for medical video generation that replaces quadratic self-attention with three cascaded attention modules—spatial, temporal, and channel—each computed in linear time. The paper's central claim is that modeling dependencies on all three dimensions, and not just space and time, is what produces realistic endoscopy videos, and that the linear design makes this affordable. On 16-frame benchmarks, the small FEAT-S variant reaches quality comparable to the state-of-the-art Endora while using only about 23% of its parameters, and the larger FEAT-L surpasses all compared methods on the reported metrics. If correct, medical video synthesis becomes cheaper to train and deploy, and generated videos are useful enough as unlabeled data to improve downstream semi-supervised disease diagnosis.

What carries the argument

The load-bearing mechanism is the weighted key-value (WKV) attention block, a linear-time attention variant, paired with global channel attention that produces a channel-interaction matrix over the spatial and temporal positions. To keep locality, FEAT replaces the original token shift in WKV with 2D depthwise convolution for the spatial dimension and 1D depthwise convolution for the temporal dimension. The three block types are stacked in a sequential spatial-temporal-channel arrangement, and ResVGM adds the input embedding as a residual to the value stream of every block with two learned per-channel scaling parameters. This combination is what lets the paper claim global dependencies across all dimensions at linear computational cost.

What would settle it

Train the same FEAT-S configuration with the original Vision-RWKV token shift in place of the 2D/1D depthwise-convolution shifts, keeping data and hyperparameters fixed; if FVD does not worsen substantially from the reported 415.4 on Colonoscopic, the claimed contribution of the shift modifications is not supported. Alternatively, measure per-token cost as the frame count grows from 16 to 32 at fixed resolution: a superlinear increase would contradict the linear-complexity claim.

Watch

Extended reading notes

Core claim

The discovery is an architecture, not a single theorem. FEAT processes latent video tokens with interleaved spatial, temporal, and channel Transformer blocks: the spatial block uses weighted key-value (WKV) attention with a 2D depthwise-convolution token shift, the temporal block uses WKV attention with a 1D depthwise-convolution token shift, and the channel block uses global channel attention. A residual value guidance module (ResVGM) injects the input embedding—which encodes both video content and the current noise pattern—as a residual value into every attention block, giving fine-grained, pixel-level denoising guidance in place of coarse timestep embeddings alone. The paper reports that every added component improves FVD monotonically (990.0, 788.4, 583.6, 415.4) on Colonoscopic, and that FEAT-S matches or slightly beats Endora on the tested metrics while FEAT-L leads on all of them across both datasets.

Load-bearing premise

The argument presumes that the borrowed WKV attention, after the 2D and 1D depthwise-convolution token-shift modifications, still gives a genuinely global spatial and temporal receptive field at linear cost on 16-frame latent video tokens; if these shifts shrink the effective receptive field, the efficiency-quality comparison loses its stated mechanism.

Editorial extensions

If this is right

  • Attention cost grows linearly with the number of video tokens, so longer or higher-resolution medical videos become more practical than with quadratic self-attention.
  • FEAT-S uses 158.0M parameters and 118.7 GFLOPs versus 673.7M and 465.8 GFLOPs for Endora, making the smaller variant materially cheaper while keeping comparable quality.
  • FEAT-generated colonoscopy videos transfer to downstream diagnosis: semi-supervised polyp F1 rises to 91.3 with FEAT-L, above Endora's 87.0 and a supervised-only 74.5 baseline.
  • Each proposed component contributes independently, since adding WKV attention, channel attention, and ResVGM in turn lowers FVD from 990.0 to 415.4 on Colonoscopic.
  • Channel interactions are no longer neglected in the transformer design, addressing a limitation shared by previous spatial-temporal video transformers.

Reading between the lines

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

  • An untested extension: because the spatial-temporal-channel decomposition is not specific to endoscopy, the same block design could presumably transfer to other medical video modalities such as ultrasound or cardiac MRI, but FEAT's experiments only cover gastrointestinal datasets.
  • A testable consequence: the residual value guidance mechanism is generic, so adding the input embedding as a residual value to any video-diffusion transformer should improve generation if coarse timestep guidance is the limiting factor; FEAT does not evaluate this outside its own architecture.
  • A caveat about the efficiency claim: the linear-complexity argument is asymptotic, and the practical speedup at the fixed 16-frame, 128x128 evaluation size would need wall-clock measurements rather than the reported parameter and FLOP counts alone.
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 / 6 minor

Summary. The paper proposes FEAT, a latent diffusion transformer for medical video generation that applies sequential spatial, temporal, and channel attention. Spatial and temporal blocks use WKV attention with modified token-shift mechanisms, the channel block uses global channel attention from Restormer, and a residual value guidance module (ResVGM) injects the input embedding as fine-grained denoising guidance. Experiments are reported on Colonoscopic and Kvasir-Capsule datasets, comparing against StyleGAN-V, LVDM, MoStGAN-V, and Endora on FVD, CD-FVD, FID, and IS, plus a semi-supervised classification downstream task and an ablation study. The headline claim is that FEAT-S, with 158.0M parameters (23% of Endora's 673.7M), achieves comparable or superior performance, while FEAT-L surpasses all evaluated baselines.

Significance. If the empirical results are robust, FEAT is a genuinely parameter-efficient architecture for medical video generation: the ablation in Table 3 is internally consistent and shows each proposed component improving FVD/FID, and the parameter-ratio arithmetic (158.0/673.7 ≈ 23%) is correct. The paper also provides code and uses public datasets, which aids reproducibility. However, the central empirical claim is not yet established because the comparisons rest on single-run metrics without uncertainty quantification or a fully specified evaluation protocol; several metric differences are small relative to expected FVD noise. The efficient-attention complexity statement also needs precision before the efficiency claim can be fully assessed.

major comments (5)
  1. [§3.1, Table 1] The central comparison is based on single-run metrics with no error bars, confidence intervals, or significance tests. On Kvasir-Capsule, FEAT-S and Endora differ by only 0.1 in FVD (72.2 vs 72.3) and 0.64 in FID (9.97 vs 10.61); on Colonoscopic, the FID difference is 0.1 (13.3 vs 13.4). These differences are likely within metric noise, so the claim that FEAT-S is "comparable or superior" with 23% of Endora's parameters is not yet supported. Please report mean and standard deviation over at least three seeds, and include a significance test (e.g., a paired bootstrap or repeated-measures test) for the headline comparisons.
  2. [§3.2] The paper states that baselines were "replicated" but provides no training details: no number of training iterations or epochs, no batch size, no learning-rate schedule, no number of GPUs, no sampling steps, and no classifier-free guidance settings. Without these, FEAT-L's larger FVD gains (e.g., 351.1 vs 460.7 on Colonoscopic) could stem from protocol mismatch rather than architectural superiority. Please report the full training and evaluation protocol for every method, including the exact FVD/CD-FVD implementation and how the 2048 evaluation samples are drawn from each dataset.
  3. [Abstract and Table 1] The claim that FEAT-L "surpasses all comparison methods" is narrower than stated because EndoMamba [10], an endoscopic video foundation model, is cited in the references but never included in the comparisons. To support a state-of-the-art claim over endoscopic video generation, either add EndoMamba to Table 1 or explicitly qualify the claim as "all evaluated baselines." In addition, Table 1 leaves parameter and FLOPs entries blank for StyleGAN-V, LVDM, and MoStGAN-V; these should be filled in or marked with a clear explanation.
  4. [§2.2, Figure 2] The paper describes the attention mechanisms as achieving "global dependencies" with "linear computational complexity" in each dimension, but this is imprecise. WKV attention has linear complexity in sequence length but is a recurrent/decay-based formulation rather than full pairwise attention, so its notion of global dependency should be stated explicitly. Global channel attention from [11] has cost O(F·H·W·C²), which is quadratic in channel count C. Please provide the explicit complexity in terms of F, H, W, and C for each of the three blocks, and clarify in what sense WKV attention is global for the spatial and temporal dimensions.
  5. [§2.3, Eq. (2)] ResVGM is a central contribution and the Table 3 ablation attributes a large improvement to it (FVD 583.6 → 415.4), yet the module is underspecified. The paper says Z is obtained "via convolution of the input" without giving the convolution kernel size, stride, or normalization, and it does not state whether the same Z is reused in all 27 blocks or whether λ1_c, λ2_c are per-block. Please clarify the computation of Z, the sharing pattern across blocks, and the initialization of the learnable scaling parameters.
minor comments (6)
  1. [References] Reference [26] is titled "HyperKvasir," but the text refers to the Kvasir-Capsule dataset; please verify that the citation is correct or add the proper Kvasir-Capsule reference.
  2. [Acknowledgments] There are typos in the acknowledgments: "Proiect" should be "Project" and "Bejing" should be "Beijing."
  3. [Table 1] The formatting of the missing FLOPs/parameter entries uses backslashes ("\ \"), which will not render correctly; please use an em dash and add a table note explaining why these values are omitted.
  4. [References] Reference [16] lists the arXiv ID as "2404.058923," which appears to have an extra digit; reference [27] lists "2104.112225(14), 6," which appears malformed. Please check both.
  5. [Figure 4] The qualitative comparison is at low resolution and does not clearly show temporal consistency; consider adding frame-by-frame zoomed crops or a temporal-structure visualization.
  6. [§3.1] The evaluation description states that FVD follows StyleGAN-V, but the CD-FVD computation is not described; please specify the official implementation used and any preprocessing differences between FVD and CD-FVD.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: FEAT's claims are empirical evaluations against external baselines on public datasets; the only self-citation is motivational and not load-bearing.

full rationale

I find no circular step in the paper's derivation chain. The headline claims (FEAT-S comparable to Endora with 23% parameters; FEAT-L surpassing comparison methods) are empirical outcomes of training on public Colonoscopic and Kvasir-Capsule datasets and measuring FVD, CD-FVD, FID, and IS; no parameter is fitted to those metrics, and no evaluation quantity is defined in terms of the architecture's outputs. The WKV/channel-attention building blocks are imported from external prior work (Vision-RWKV [17], Restormer [11]), and the linear-complexity statement is cited from [19]; the 2D/1D depth-wise convolution token shifts are a design choice, not an input to the result. Section 2.3's Eq. (2) defines the ResVGM output as a function of the input embedding Z, but this is an architectural definition, not a predicted quantity derived from itself. The one self-citation ([18], Restore-RWKV, overlapping authors) appears only as motivational support for 'exceptional performance of weighted key-value (WKV) attention ... in denoising' and is paired with external citation [17]; it is not the basis for the Table 1 numbers or the downstream F1 improvements. Section 4's statement that 'Future work will extend FEAT to additional imaging modalities and conduct more comprehensive evaluations [30]' is a limitation of evaluation breadth, not a circularity. The absence of seeds, confidence intervals, and baseline training details is a statistical robustness concern, distinct from circularity. Overall, the paper is self-contained against external benchmarks, and the central claims do not reduce to the self-citation or to fitted inputs.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no derived constants or fitted scientific parameters; its free parameters are standard neural network weights and hand-chosen architectural sizes. The central claims rest on imported prior results (WKV attention, channel attention, latent VAE, evaluation metrics) rather than on new derivations, so the assumption ledger contains these background commitments.

free parameters (3)
  • Hidden dimension d = 512 for FEAT-S, 1024 for FEAT-L
    Hand-chosen model width separating small and large variants; no search or derivation reported.
  • Block count per dimension = 9 spatial, 9 temporal, 9 channel (27 blocks)
    Architecture design choice in Section 3.1; alternative interleavings are not ablated.
  • Residual guidance weights lambda1_c, lambda2_c = learned per-channel vectors
    Trainable parameters in Eq. (2); their values are fit by training, not derived.
assumptions (5)
  • domain assumption Stable Diffusion VAE produces faithful latent representations for 16-frame medical videos
    Section 2.1 adopts latent diffusion with a pretrained VAE [21,33]; if VAE reconstruction distorts endoscopic texture, FVD/FID gains would not reflect true video quality.
  • domain assumption WKV attention and modified token shifts give global spatial-temporal attention with linear complexity
    Section 2.2 borrows WKV from Vision-RWKV [17] and alters the token shift with depth-wise convolutions; no derivation or ablation shows these modifications preserve global dependency and linear cost in video tokens.
  • domain assumption Global channel attention from Restormer [11] improves denoising in diffusion transformers
    Section 2.2 directly employs the channel attention block [11]; its benefit is shown only via the aggregate ablation, not isolated for channel attention alone.
  • domain assumption FVD and CD-FVD computed with StyleGAN-V's 2048-sample protocol reliably rank video generators
    Section 3.1 specifies metrics from prior work; validity of ranking depends on these metrics and the fixed evaluation setting.
  • domain assumption FixMatch semi-supervised learning with generated videos improves downstream polyp diagnosis
    Section 3.3 uses FixMatch [36]; results assume the framework and the domain match between generated and real data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FEAT: Full-Dimensional Efficient Attention Transformer for Medical Video Generation." pith.science (2026). https://pith.science/paper/BXXSNT4R

@misc{pith2026250604956,
  author       = {Pith},
  title        = {Pith review of: FEAT: Full-Dimensional Efficient Attention Transformer for Medical Video Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BXXSNT4R}},
  note         = {Machine review of arXiv:2506.04956}
}
read the original abstract

Synthesizing high-quality dynamic medical videos remains a significant challenge due to the need for modeling both spatial consistency and temporal dynamics. Existing Transformer-based approaches face critical limitations, including insufficient channel interactions, high computational complexity from self-attention, and coarse denoising guidance from timestep embeddings when handling varying noise levels. In this work, we propose FEAT, a full-dimensional efficient attention Transformer, which addresses these issues through three key innovations: (1) a unified paradigm with sequential spatial-temporal-channel attention mechanisms to capture global dependencies across all dimensions, (2) a linear-complexity design for attention mechanisms in each dimension, utilizing weighted key-value attention and global channel attention, and (3) a residual value guidance module that provides fine-grained pixel-level guidance to adapt to different noise levels. We evaluate FEAT on standard benchmarks and downstream tasks, demonstrating that FEAT-S, with only 23\% of the parameters of the state-of-the-art model Endora, achieves comparable or even superior performance. Furthermore, FEAT-L surpasses all comparison methods across multiple datasets, showcasing both superior effectiveness and scalability. Code is available at https://github.com/Yaziwel/FEAT.

Figures

Figures reproduced from arXiv: 2506.04956 by the authors.

Figure 1
Figure 1. The pipeline of FEAT for medical video generation. (a) Architecture of conven￾tional models using cascaded spatial-temporal Transformer blocks. (b) Architecture of FEAT, which incorporates cascaded spatial-temporal-channel Transformer blocks. (c) Details of the conventional Transformer block, featuring quadratic computational com￾plexity for self-attention and global timestep guidance. (d) Details of the Transformer… view at source ↗
Figure 2
Figure 2. Three distinct Transformer blocks in FEAT. (a) Spatial Transformer block with WKV attention [17]. (b) Temporal Transformer block with WKV attention [17]. (c) Channel Transformer block with global channel attention [11]. F, H, W, and C represent the frame number, height, width, and channel of the input feature, respectively. (d)). (3) To address the coarse, global-level guidance that struggles to adapt to varying noi… view at source ↗
Figure 3
Figure 3. The schematic diagram of the proposed ResVGM, where (a) represents the original framework, and (b) denotes the framework with ResVGM incorporated to dif￾ferent Transformer blocks. The frameworks primarily illustrate operations surrounding the attention mechanism in Transformer blocks, where ResVGM is integrated, while other modules are omitted for simplicity. content-dependent denoising at specific noise levels. As … view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Qualitative Comparison on Colonoscopic and Kvasir-Capsule Datasets. video samples, with each sample maintaining the complete 16-frame temporal structure to preserve motion dynamics and temporal coherence. Implementation Details. Our implementation employs the AdamW opt…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Med-URWKV{\dag}: Toward Enhanced Pretrained Pure VRWKV Models for Medical Image Segmentation

    eess.IV 2025-06 conditional novelty 5.0 of 10

    Pretrained pure VRWKV encoders paired with pure VRWKV decoders match or beat CNN, ViT, and Mamba baselines, with a small model plus FAWA and MSCF modules reaching 88% average Dice.

Reference graph

Works this paper leans on

36 extracted references · 10 canonical work pages · cited by 1 Pith paper

  1. [10]

    Tian, Q., Liao, H., Huang, X., Yang, B., Lei, D., Ourselin, S., Liu, H.: Endomamba: An efficient foundation model for endoscopic videos (2025), https://arxiv.org/abs/ 2502.19090

    work page Pith review arXiv 2025

  2. [11]

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

    Zamir, S.W., Arora, A., Khan, S., Hayat, M., Khan, F.S., Yang, M.H.: Restormer: Efficient transformer for high-resolution image restoration. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5728–5739 (2022)

    2022

  3. [1]

    In: Medical imaging with deep learning (2022)

    Dorjsembe, Z., Odonchimed, S., Xiao, F.: Three-dimensional medical image syn- thesis with denoising diffusion probabilistic models. In: Medical imaging with deep learning (2022)

    2022

  4. [2]

    Soft Masked Mamba Diffusion Model for CT to MRI Conversion

    Wang, Z., Zhang, L., Wang, L., Zhang, Z.: Soft masked mamba diffusion model for ct to mri conversion. arXiv preprint arXiv:2406.15910 (2024)

    work page Pith review arXiv 2024

  5. [3]

    In: Proceedings of the IEEE/CVF International conference on com- puter vision

    Liu, J., Anirudh, R., Thiagarajan, J.J., He, S., Mohan, K.A., Kamilov, U.S., Kim, H.: Dolce: A model-based probabilistic diffusion framework for limited-angle ct re- construction. In: Proceedings of the IEEE/CVF International conference on com- puter vision. pp. 10498–10508 (2023)

    work page 2023

  6. [4]

    arXiv preprint arXiv:2209.14792 (2022)

    Singer, U., Polyak, A., Hayes, T., Yin, X., An, J., Zhang, S., Hu, Q., Yang, H., Ashual, O., Gafni, O., et al.: Make-a-video: Text-to-video generation without text- video data. arXiv preprint arXiv:2209.14792 (2022)

    arXiv 2022

  7. [5]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Khachatryan, L., Movsisyan, A., Tadevosyan, V., Henschel, R., Wang, Z., Navasardyan, S., Shi, H.: Text2video-zero: Text-to-image diffusion models are zero- shot video generators. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 15954–15964 (2023)

    work page 2023

  8. [6]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Chen, H., Zhang, Y., Cun, X., Xia, M., Wang, X., Weng, C., Shan, Y.: Videocrafter2: Overcoming data limitations for high-quality video diffusion mod- els. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7310–7320 (2024) 10 Wang, Yang et al

    work page 2024

Show all 36 references
  1. [7]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Li, C., Liu, H., Liu, Y., Feng, B.Y., Li, W., Liu, X., Chen, Z., Shao, J., Yuan, Y.: Endora: Video generation models as endoscopy simulators. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 230–240. Springer (2024)

    2024

  2. [8]

    IEEE Transactions on Visualization and Computer Graphics (2024)

    Xing, J., Xia, M., Liu, Y., Zhang, Y., Zhang, Y., He, Y., Liu, H., Chen, H., Cun, X., Wang, X., et al.: Make-your-video: Customized video generation using textual and structural guidance. IEEE Transactions on Visualization and Computer Graphics (2024)

    2024

  3. [9]

    arXiv preprint arXiv:2401.03048 (2024)

    Ma, X., Wang, Y., Jia, G., Chen, X., Liu, Z., Li, Y.F., Chen, C., Qiao, Y.: Latte: Latent diffusion transformer for video generation. arXiv preprint arXiv:2401.03048 (2024)

    2024 arXiv

  4. [12]

    Peebles,W.,Xie,S.:Scalablediffusionmodelswithtransformers.In:Proceedingsof the IEEE/CVF international conference on computer vision. pp. 4195–4205 (2023)

    2023

  5. [13]

    arXiv preprint arXiv:2411.01647 (2024)

    Wang, Z., Zhang, L., Wang, L., Zhu, M., Zhang, Z.: Optical flow representation alignment mamba diffusion model for medical video generation. arXiv preprint arXiv:2411.01647 (2024)

    2024 arXiv

  6. [14]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P., Joulin, A.: Emerging properties in self-supervised vision transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 9650–9660 (2021)

    2021

  7. [15]

    arXiv preprint arXiv:2305.13048 (2023)

    Peng, B., Alcaide, E., Anthony, Q., Albalak, A., Arcadinho, S., Biderman, S., Cao, H., Cheng, X., Chung, M., Grella, M., et al.: Rwkv: Reinventing rnns for the transformer era. arXiv preprint arXiv:2305.13048 (2023)

    2023 arXiv

  8. [16]

    arXiv preprint arXiv:2404.058923 (2024)

    Peng, B., Goldstein, D., Anthony, Q., Albalak, A., Alcaide, E., Biderman, S., Cheah, E., Ferdinan, T., Hou, H., Kazienko, P., et al.: Eagle and finch: Rwkv with matrix-valued states and dynamic recurrence. arXiv preprint arXiv:2404.058923 (2024)

    2024 arXiv

  9. [17]

    arXiv preprint arXiv:2403.02308 (2024)

    Duan, Y., Wang, W., Chen, Z., Zhu, X., Lu, L., Lu, T., Qiao, Y., Li, H., Dai, J., Wang, W.: Vision-rwkv: Efficient and scalable visual perception with rwkv-like architectures. arXiv preprint arXiv:2403.02308 (2024)

    2024 arXiv

  10. [18]

    arXiv preprint arXiv:2407.11087 (2024)

    Yang, Z., Li, J., Zhang, H., Zhao, D., Wei, B., Xu, Y.: Restore-rwkv: Efficient and effective medical image restoration with rwkv. arXiv preprint arXiv:2407.11087 (2024)

    2024 arXiv

  11. [19]

    In: Proceedings of the IEEE/CVF winter conference on appli- cations of computer vision

    Shen, Z., Zhang, M., Zhao, H., Yi, S., Li, H.: Efficient attention: Attention with linear complexities. In: Proceedings of the IEEE/CVF winter conference on appli- cations of computer vision. pp. 3531–3539 (2021)

    2021

  12. [20]

    Advances in neural information processing systems34, 21696–21707 (2021)

    Kingma, D., Salimans, T., Poole, B., Ho, J.: Variational diffusion models. Advances in neural information processing systems34, 21696–21707 (2021)

    2021

  13. [21]

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

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022)

    2022

  14. [22]

    arXiv preprint arXiv:2211.13221 (2022) FEAT: Full-Dimensional Efficient Attention Transformer 11

    He, Y., Yang, T., Zhang, Y., Shan, Y., Chen, Q.: Latent video diffusion models for high-fidelity long video generation. arXiv preprint arXiv:2211.13221 (2022) FEAT: Full-Dimensional Efficient Attention Transformer 11

    2022 arXiv

  15. [23]

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

    Chollet, F.: Xception: Deep learning with depthwise separable convolutions. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1251–1258 (2017)

    2017

  16. [24]

    arXiv preprint arXiv:2410.17897 (2024)

    Zhou,Z.,Wu,T.,Jiang,Z.,Lan,Z.:Valueresiduallearningforalleviatingattention concentration in transformers. arXiv preprint arXiv:2410.17897 (2024)

    2024 arXiv

  17. [25]

    IEEE transactions on medical imaging35(9), 2051–2063 (2016)

    Mesejo, P., Pizarro, D., Abergel, A., Rouquette, O., Beorchia, S., Poincloux, L., Bartoli, A.: Computer-aided classification of gastrointestinal lesions in regular colonoscopy. IEEE transactions on medical imaging35(9), 2051–2063 (2016)

    2016

  18. [26]

    Sci- entific data7(1), 283 (2020)

    Borgli, H., Thambawita, V., Smedsrud, P.H., Hicks, S., Jha, D., Eskeland, S.L., Randel, K.R., Pogorelov, K., Lux, M., Nguyen, D.T.D., et al.: Hyperkvasir, a com- prehensive multi-class image and video dataset for gastrointestinal endoscopy. Sci- entific data7(1), 283 (2020)

    2020

  19. [27]

    arXiv preprint arXiv:2104.112225(14), 6 (2021)

    Parmar, G., Zhang, R., Zhu, J.Y.: On buggy resizing libraries and surprising sub- tleties in fid calculation. arXiv preprint arXiv:2104.112225(14), 6 (2021)

    2021 arXiv

  20. [28]

    In: Proceedings of the IEEE international conference on computer vision

    Saito, M., Matsumoto, E., Saito, S.: Temporal generative adversarial nets with singular value clipping. In: Proceedings of the IEEE international conference on computer vision. pp. 2830–2839 (2017)

    2017

  21. [29]

    arXiv preprint arXiv:1812.01717 (2018)

    Unterthiner, T., Van Steenkiste, S., Kurach, K., Marinier, R., Michalski, M., Gelly, S.: Towards accurate generative models of video: A new metric & challenges. arXiv preprint arXiv:1812.01717 (2018)

    2018 arXiv

  22. [30]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ge, S., Mahapatra, A., Parmar, G., Zhu, J.Y., Huang, J.B.: On the content bias in fréchet video distance. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7277–7288 (2024)

    2024

  23. [31]

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

    Skorokhodov, I., Tulyakov, S., Elhoseiny, M.: Stylegan-v: A continuous video gen- erator with the price, image quality and perks of stylegan2. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 3626–3636 (2022)

    2022

  24. [32]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Shen, X., Li, X., Elhoseiny, M.: Mostgan-v: Video generation with temporal motion styles. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5652–5661 (2023)

    2023

  25. [33]

    arXiv preprint arXiv:2311.15127 (2023)

    Blattmann, A., Dockhorn, T., Kulal, S., Mendelevitch, D., Kilian, M., Lorenz, D., Levi, Y., English, Z., Voleti, V., Letts, A., et al.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023)

    2023 arXiv

  26. [34]

    In: 2021 IEEE International Conference on Image Pro- cessing (ICIP)

    Li, C., Zhang, Y., Liang, Z., Ma, W., Huang, Y., Ding, X.: Consistent poste- rior distributions under vessel-mixing: a regularization for cross-domain retinal artery/vein classification. In: 2021 IEEE International Conference on Image Pro- cessing (ICIP). pp. 61–65. IEEE (2021)

    2021

  27. [35]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Tian, Y., Pang, G., Liu, F., Liu, Y., Wang, C., Chen, Y., Verjans, J., Carneiro, G.: Contrastive transformer-based multiple instance learning for weakly supervised polyp frame detection. In: International Conference on Medical Image Computing and Computer-Assisted Intervention...

    2022

  28. [36]

    Advances in neural information processing systems33, 596–608 (2020)

    Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C.A., Cubuk, E.D., Kurakin, A., Li, C.L.: Fixmatch: Simplifying semi-supervised learning with consistency and confidence. Advances in neural information processing systems33, 596–608 (2020)

    2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.