Pith. sign in

REVIEW 4 major objections 4 minor 50 references

Temporal Cluster Assignment for Efficient Real-Time Video Segmentation

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read TCA shows that temporal coherence can refine token clusters across frames, giving clustering-based video segmentation a free accuracy-speed boost without fine-tuning.

desk verdict A plausible temporal token-clustering idea worth a referee, but the submitted text is mojibake so its numbers cannot be checked. read the letter →

arxiv 2508.05851 v1 pith:KMYLMJKJ submitted 2025-08-07 cs.CV

Ka-Wai Yung , Felix J. S. Bragman , Jialang Xu , Imanol Luengo , Danail Stoyanov , Evangelos B. Mazomenos This is my paper
classification cs.CV
keywords videosegmentationtokenclusteringtemporalcoherencefine-tuning-freeSwinTransformerreal-timeinferenceYouTube-VISOVIS
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 introduces Temporal Cluster Assignment (TCA), a lightweight, fine-tuning-free strategy for clustering-based video segmentation. Its central claim is that temporal coherence is a usable signal for token redundancy: cluster assignments that work for one frame remain informative for the next, so the model can refine rather than recompute them, preserving fine-grained details while cutting computation. If correct, Swin-based real-time video segmentation becomes faster at equal or higher accuracy, and the benefit transfers from natural videos to a private surgical video dataset without retraining.

What carries the argument

Temporal Cluster Assignment (TCA): a training-free refinement step that carries token cluster assignments across frames and updates them using temporal correlations. It sidesteps the fixed tokens-per-window constraint of Swin Transformer attention while exploiting the redundancy between consecutive frames, so the model spends computation on genuinely changed details rather than recomputing stable clusters.

What would settle it

Run TCA against per-frame clustering on video clips with abrupt scene cuts, fast camera motion, or rapid instrument motion in the surgical domain and compare accuracy against latency; if the accuracy-speed trade-off does not improve on those clips, the temporal-coherence premise is falsified.

Watch

Extended reading notes

Core claim

The paper claims that temporal correlations across frames can be exploited to improve token clustering for video segmentation without any fine-tuning. Instead of pruning redundant tokens independently per frame, TCA refines token clusters using information from temporally adjacent frames, keeping fine-grained details while reducing compute. Evaluated on YouTube-VIS 2019, YouTube-VIS 2021, OVIS, and a private surgical video dataset, the method consistently improves the accuracy-speed trade-off of existing clustering-based methods, and the authors report that it generalizes across both natural and domain-specific videos.

Load-bearing premise

The claim depends on cluster assignments from one frame still being informative for the next frame; if motion, occlusion, or domain shifts make temporal correlation an unreliable proxy for token redundancy, the accuracy-speed improvement could shrink or disappear.

Editorial extensions

If this is right

  • Clustering-based video segmentation models can obtain a consistent accuracy-speed improvement by adding TCA, with no fine-tuning or model weight changes.
  • Real-time deployment of larger Swin-based segmentation models becomes more practical on resource-constrained hardware.
  • Temporal redundancy can be exploited even when window attention requires a fixed number of tokens per window.
  • The reported gains on a private surgical video dataset suggest the approach transfers to domain-specific videos beyond natural scenes.

Reading between the lines

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

  • A natural stress test is fast camera motion or abrupt scene cuts, where temporal correlation weakens; a per-frame similarity gate could decide when to fall back to recomputing clusters.
  • The same cluster-refinement idea could extend beyond Swin to other window-based or hierarchical transformers used for dense prediction tasks such as detection or panoptic segmentation.
  • Because TCA is fine-tuning-free, it could be combined with pruning, distillation, or quantization to compound compute savings.
  • If frame-to-frame correlation is measured explicitly, TCA could become adaptive, spending less compute on near-static segments and more on fast-moving ones.
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

4 major / 4 minor

Summary. The paper introduces Temporal Cluster Assignment (TCA), a fine-tuning-free method that augments token clustering with temporal correlations across frames for Swin-based video segmentation. TCA is intended to reduce computational cost while preserving fine-grained details, and the abstract claims that it consistently improves the accuracy--speed trade-off of existing clustering-based methods on YouTube-VIS 2019, YouTube-VIS 2021, OVIS, and a private surgical video dataset. The abstract is the only readable portion of the submitted manuscript; the body, equations, algorithm description, and tabular results are corrupted mojibake. Consequently, the method details and the experimental evidence behind the central claim could not be inspected.

Significance. If the claimed results hold, TCA would be a practical and useful contribution: it is designed to be fine-tuning-free, compatible with Swin's fixed-window attention constraint, and applicable across natural and domain-specific videos. The choice of two public benchmarks (YouTube-VIS 2019/2021 and OVIS) is a strength, and the central claim is falsifiable once quantitative numbers are available. However, the current submission provides no accessible numerical evidence, no ablations, and no reproduction artifacts; the private surgical dataset further limits independent verification. The temporal-coherence premise is plausible but untested under fast motion, occlusion, or domain shift. As submitted, the significance of the work cannot be assessed.

major comments (4)
  1. [Full text (all sections after the Abstract)] The submitted body is mojibake: equations, the algorithm description, experimental setup, and all tables are unreadable. The paper's central claim is an empirical accuracy--speed improvement, and the tables are the load-bearing evidence. I cannot verify the claimed improvements, the baselines, the metrics, or the ablations. This is not a cosmetic issue; it prevents substantive review. A correctly encoded PDF must be provided before the manuscript can be evaluated.
  2. [Abstract] The abstract states that TCA 'consistently boosts the accuracy-speed trade-off' on four datasets, but reports no numerical results, no error bars, no comparison magnitudes, and no ablation. 'Consistently' is unsupported without quantitative effect sizes. Even if the body were readable, the abstract should state representative accuracy and speed deltas (e.g., mask AP or J&F versus FPS) and the number of runs or variance.
  3. [Method (TCA temporal-correlation step)] The method's core premise is that token clusters from the previous frame remain informative for the current frame. Under fast camera or object motion, occlusion, or domain shift, stale cluster assignments could misassign boundary tokens and degrade accuracy. No theoretical argument or targeted robustness experiment is visible in the unreadable text. Please add an analysis of performance as a function of inter-frame displacement or occlusion, or clearly state the conditions under which TCA is expected to fail.
  4. [Datasets (private surgical video dataset)] One of the four evaluation domains is a private surgical video dataset. The claim that TCA 'generalizes competently across both natural and domain-specific videos' cannot be independently checked without a public protocol: dataset size, annotation procedure, metrics, and ideally de-identified examples or a data-availability statement. If this information exists in the body, it is currently unreadable.
minor comments (4)
  1. [Abstract] Include concrete numbers and error bars for the headline accuracy--speed trade-off claim.
  2. [Terminology] Define 'fine-tuning-free' operationally: no learned parameters are modified, but are the clustering hyperparameters (cluster count, temporal window size, refinement threshold) tuned on the evaluation sets? This affects the strength of the claim.
  3. [Related Work / References] The corrupted text prevented checking the references to token-clustering methods and Swin variants. Please verify all citations in the resubmitted manuscript.
  4. [Wording] The phrase 'generalizes competently' is vague; replace it with a quantified generalization gap or per-dataset accuracy--speed comparisons.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity identified: the central claim is supported by external benchmarks and no equation-level reduction or fitted-prediction equivalence is present in the readable text.

full rationale

TCA's claim is that exploiting temporal coherence improves the accuracy/speed trade-off of clustering-based video segmentation. The abstract grounds this in evaluations on YouTube-VIS 2019, YouTube-VIS 2021, OVIS, and a private surgical video dataset, i.e., external datasets rather than quantities fitted inside the method. The method statement 'TCA refines token clusters using temporal correlations' is a design premise, not a restatement of the target accuracy/speed result. The supplied full text is largely mojibake, so no equation-level derivation chain can be inspected and no 'Eq. X = Eq. Y by construction' reduction can be exhibited. The private surgical dataset is not externally auditable, which is a verification gap but not circularity. No self-citation chain, no imported uniqueness theorem, and no fitted parameter renamed as a prediction appear in the readable portions. Under the hard rule requiring quotation and demonstration of a specific reduction, no circular step is found.

Assumptions & free parameters 1 free parameters · 2 assumptions · 0 invented entities

Everything here rests on abstract-level statements because the full text is unreadable. The only free parameter is the set of clustering and refinement settings, whose values are not visible. No new physical entities are introduced; TCA is an algorithmic strategy, not a new force, particle, or conserved quantity.

free parameters (1)
  • TCA clustering hyperparameters (e.g., cluster count, temporal window size, refinement threshold)
    The abstract mentions no numeric settings; the full text is corrupted, so any values chosen by the authors cannot be audited. Token clustering methods typically carry at least one such knob.
assumptions (2)
  • domain assumption Neighboring frames share enough visual structure that reusing cluster assignments across frames preserves segmentation quality.
    Invoked in the abstract's method description; it is the core premise that makes temporal refinement safe.
  • domain assumption Swin window attention requires a fixed number of tokens per window, so token reduction must preserve window consistency.
    Stated in the abstract as the reason conventional pruning does not apply directly to Swin.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Temporal Cluster Assignment for Efficient Real-Time Video Segmentation." pith.science (2026). https://pith.science/paper/KMYLMJKJ

@misc{pith2026250805851,
  author       = {Pith},
  title        = {Pith review of: Temporal Cluster Assignment for Efficient Real-Time Video Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KMYLMJKJ}},
  note         = {Machine review of arXiv:2508.05851}
}
read the original abstract

Vision Transformers have substantially advanced the capabilities of segmentation models across both image and video domains. Among them, the Swin Transformer stands out for its ability to capture hierarchical, multi-scale representations, making it a popular backbone for segmentation in videos. However, despite its window-attention scheme, it still incurs a high computational cost, especially in larger variants commonly used for dense prediction in videos. This remains a major bottleneck for real-time, resource-constrained applications. Whilst token reduction methods have been proposed to alleviate this, the window-based attention mechanism of Swin requires a fixed number of tokens per window, limiting the applicability of conventional pruning techniques. Meanwhile, training-free token clustering approaches have shown promise in image segmentation while maintaining window consistency. Nevertheless, they fail to exploit temporal redundancy, missing a key opportunity to further optimize video segmentation performance. We introduce Temporal Cluster Assignment (TCA), a lightweight and effective, fine-tuning-free strategy that enhances token clustering by leveraging temporal coherence across frames. Instead of indiscriminately dropping redundant tokens, TCA refines token clusters using temporal correlations, thereby retaining fine-grained details while significantly reducing computation. Extensive evaluations on YouTube-VIS 2019, YouTube-VIS 2021, OVIS, and a private surgical video dataset show that TCA consistently boosts the accuracy-speed trade-off of existing clustering-based methods. Our results demonstrate that TCA generalizes competently across both natural and domain-specific videos.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 40 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Athar, A.; Hermans, A.; Luiten, J.; Ramanan, D.; and Leibe, B. 2023. Tarvis: A unified approach for target-based video segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18738--18748

    work page 2023

  4. [4]

    Bala z evi \'c , I.; Shi, Y.; Papalampidi, P.; Chaabouni, R.; Koppula, S.; and H \'e naff, O. J. 2024. Memory consolidation enables long-context video understanding. arXiv preprint arXiv:2402.05861

    arXiv 2024

  5. [5]

    Bolya, D.; Fu, C.-Y.; Dai, X.; Zhang, P.; Feichtenhofer, C.; and Hoffman, J. 2022. Token Merging: Your ViT but faster. arXiv preprint arXiv:2210.09461

    arXiv 2022

  6. [6]

    Cheng, B.; Choudhuri, A.; Misra, I.; Kirillov, A.; Girdhar, R.; and Schwing, A. G. 2021. Mask2former for video instance segmentation. arXiv preprint arXiv:2112.10764

    arXiv 2021

  7. [7]

    G.; Kirillov, A.; and Girdhar, R

    Cheng, B.; Misra, I.; Schwing, A. G.; Kirillov, A.; and Girdhar, R. 2022. Masked-Attention Mask Transformer for Universal Image Segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1290--1299

    work page 2022

  8. [8]

    Dutson, M.; Li, Y.; and Gupta, M. 2023. Eventful transformers: Leveraging temporal redundancy in vision transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 16911--16923

    work page 2023

Show all 50 references
  1. [9]

    A.; Jafari, F

    Fayyaz, M.; Koohpayegani, S. A.; Jafari, F. R.; Sengupta, S.; Joze, H. R. V.; Sommerlade, E.; Pirsiavash, H.; and Gall, J. 2022. Adaptive token sampling for efficient vision transformers. In European Conference on Computer Vision, 396--414. Springer

    2022

  2. [10]

    Geiger, A.; Lenz, P.; and Urtasun, R. 2012. Are we ready for autonomous driving? the kitti vision benchmark suite. In 2012 IEEE conference on computer vision and pattern recognition, 3354--3361. IEEE

    2012

  3. [11]

    Guo, D.; Fan, D.-P.; Lu, T.; Sakaridis, C.; and Van Gool, L. 2024. Vanishing-point-guided video semantic segmentation of driving scenes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3544--3553

    2024

  4. [12]

    B.; Escalera, S.; Taylor, G

    Haurum, J. B.; Escalera, S.; Taylor, G. W.; and Moeslund, T. B. 2023. Which tokens to use? investigating token reduction in vision transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 773--783

    2023

  5. [13]

    W.; Lee, J.-Y.; and Kim, S

    Heo, M.; Hwang, S.; Hyun, J.; Kim, H.; Oh, S. W.; Lee, J.-Y.; and Kim, S. J. 2023. A generalized framework for video instance segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 14623--14632

    2023

  6. [14]

    W.; Lee, J.-Y.; and Kim, S

    Heo, M.; Hwang, S.; Oh, S. W.; Lee, J.-Y.; and Kim, S. J. 2022. Vita: Video instance segmentation via object token association. Advances in Neural Information Processing Systems, 35: 23109--23120

    2022

  7. [15]

    Huang, D.-A.; Yu, Z.; and Anandkumar, A. 2022. Minvis: A minimal video instance segmentation framework without video-based training. Advances in Neural Information Processing Systems, 35: 31265--31277

    2022

  8. [16]

    M.; and Habibian, A

    Kahatapitiya, K.; Karjauv, A.; Abati, D.; Porikli, F.; Asano, Y. M.; and Habibian, A. 2024. Object-centric diffusion for efficient video editing. In European Conference on Computer Vision, 91--108

    2024

  9. [17]

    P.; and Siam, M

    Karim, R.; Zhao, H.; Wildes, R. P.; and Siam, M. 2023. MED-VT: Multiscale encoder-decoder video transformer with application to object segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 6323--6333

    2023

  10. [18]

    Kim, M.; Gao, S.; Hsu, Y.-C.; Shen, Y.; and Jin, H. 2024. Token fusion: Bridging the gap between token pruning and token merging. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 1383--1392

    2024

  11. [19]

    Kong, Z.; Dong, P.; Ma, X.; Meng, X.; Niu, W.; Sun, M.; Shen, X.; Yuan, G.; Ren, B.; Tang, H.; et al. 2022. Spvit: Enabling faster vision transformers via latency-aware soft token pruning. In European Conference on Computer Vision, 620--640. Springer

    2022

  12. [20]

    Lee, S.-H.; Wang, J.; Zhang, Z.; Fan, D.; and Li, X. 2024. Video Token Merging for Long Video Understanding. In The Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  13. [21]

    Li, J.; Wang, Y.; Zhang, X.; Shi, B.; Jiang, D.; Li, C.; Dai, W.; Xiong, H.; and Tian, Q. 2023 a . AiluRus: a scalable VIT framework for dense prediction. Advances in Neural Information Processing Systems, 36: 30979--30996

    2023

  14. [22]

    Li, M.; Li, S.; Zhang, X.; and Zhang, L. 2024. Univs: Unified and universal video segmentation with prompts as queries. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3227--3238

    2024

  15. [23]

    Li, X.; Yuan, H.; Zhang, W.; Cheng, G.; Pang, J.; and Loy, C. C. 2023 b . Tube-link: A flexible cross tube framework for universal video segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 13923--13933

    2023

  16. [24]

    Li, X.; Zhang, W.; Pang, J.; Chen, K.; Cheng, G.; Tong, Y.; and Loy, C. C. 2022. Video k-net: A simple, strong, and unified baseline for video segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18847--18857

    2022

  17. [25]

    Li, Y.; Shi, J.; and Lin, D. 2018. Low-latency video semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 5997--6005

    2018

  18. [26]

    Liang, W.; Yuan, Y.; Ding, H.; Luo, X.; Lin, W.; Jia, D.; Zhang, Z.; Zhang, C.; and Hu, H. 2022 a . Expediting large-scale vision transformer for dense prediction without fine-tuning. Advances in Neural Information Processing Systems, 35: 35462--35477

    2022

  19. [27]

    Liang, Y.; Ge, C.; Tong, Z.; Song, Y.; Wang, J.; and Xie, P. 2022 b . Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations. In International Conference on Learning Representations

    2022

  20. [28]

    Liu, Y.; Gehrig, M.; Messikommer, N.; Cannici, M.; and Scaramuzza, D. 2024. Revisiting token pruning for object detection and instance segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2658--2668

    2024

  21. [29]

    Liu, Z.; Lin, Y.; Cao, Y.; Hu, H.; Wei, Y.; Zhang, Z.; Lin, S.; and Guo, B. 2021. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 10012--10022

    2021

  22. [30]

    Long, S.; Zhao, Z.; Pi, J.; Wang, S.; and Wang, J. 2023. Beyond attentive tokens: Incorporating token importance and diversity for efficient vision transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10334--10343

    2023

  23. [31]

    Luo, Z.; Xiao, Y.; Liu, Y.; Li, S.; Wang, Y.; Tang, Y.; Li, X.; and Yang, Y. 2023. Soc: Semantic-assisted object cluster for referring video object segmentation. Advances in Neural Information Processing Systems, 36: 26425--26437

    2023

  24. [32]

    Meng, L.; Li, H.; Chen, B.-C.; Lan, S.; Wu, Z.; Jiang, Y.-G.; and Lim, S.-N. 2022. Adavit: Adaptive vision transformers for efficient image recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12309--12318

    2022

  25. [33]

    Norouzi, N.; Orlova, S.; De Geus, D.; and Dubbelman, G. 2024. ALGM: Adaptive Local-then-Global Token Merging for Efficient Semantic Segmentation with Plain Vision Transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15773--15782

    2024

  26. [34]

    H.; and Bai, S

    Qi, J.; Gao, Y.; Hu, Y.; Wang, X.; Liu, X.; Bai, X.; Belongie, S.; Yuille, A.; Torr, P. H.; and Bai, S. 2022. Occluded video instance segmentation: A benchmark. International Journal of Computer Vision, 130(8): 2022--2039

    2022

  27. [35]

    Ranftl, R.; Bochkovskiy, A.; and Koltun, V. 2021. Vision transformers for dense prediction. In Proceedings of the IEEE/CVF international conference on computer vision, 12179--12188

    2021

  28. [36]

    Rao, Y.; Zhao, W.; Liu, B.; Lu, J.; Zhou, J.; and Hsieh, C.-J. 2021. Dynamicvit: Efficient vision transformers with dynamic token sparsification. Advances in neural information processing systems, 34: 13937--13949

    2021

  29. [37]

    Sarkar, S.; Datta, G.; Kundu, S.; Zheng, K.; Bhattacharyya, C.; and Beerel, P. A. 2024. MaskVD: Region Masking for Efficient Video Object Detection. arXiv preprint arXiv:2407.12067

    2024 arXiv

  30. [38]

    Song, E.; Chai, W.; Wang, G.; Zhang, Y.; Zhou, H.; Wu, F.; Chi, H.; Guo, X.; Ye, T.; Zhang, Y.; et al. 2024. Moviechat: From dense token to sparse memory for long video understanding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18221--18232

    2024

  31. [39]

    Tang, Q.; Zhang, B.; Liu, J.; Liu, F.; and Liu, Y. 2023. Dynamic token pruning in plain vision transformers for semantic segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 777--786

    2023

  32. [40]

    Tong, Z.; Song, Y.; Wang, J.; and Wang, L. 2022. Videomae: Masked autoencoders are data-efficient learners for self-supervised video pre-training. Advances in neural information processing systems, 35: 10078--10093

    2022

  33. [41]

    Wang, H.; Dedhia, B.; and Jha, N. K. 2024. Zero-TPrune: Zero-shot token pruning through leveraging of the attention graph in pre-trained transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16070--16079

    2024

  34. [42]

    Wei, S.; Ye, T.; Zhang, S.; Tang, Y.; and Liang, J. 2023. Joint token pruning and squeezing towards more aggressive compression of vision transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2092--2101

    2023

  35. [43]

    Wu, J.; Jiang, Y.; Bai, S.; Zhang, W.; and Bai, X. 2022. Seqformer: Sequential transformer for video instance segmentation. In European Conference on Computer Vision, 553--569. Springer

    2022

  36. [44]

    Wu, Q.; Yang, T.; Wu, W.; and Chan, A. B. 2023. Scalable video object segmentation with simplified framework. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 13879--13889

    2023

  37. [45]

    Xu, Y.-S.; Fu, T.-J.; Yang, H.-K.; and Lee, C.-Y. 2018. Dynamic video segmentation network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 6556--6565

    2018

  38. [46]

    Yang, L.; Fan, Y.; and Xu, N. 2019. Video instance segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 5188--5197

    2019

  39. [47]

    M.; Mallya, A.; Kautz, J.; and Molchanov, P

    Yin, H.; Vahdat, A.; Alvarez, J. M.; Mallya, A.; Kautz, J.; and Molchanov, P. 2022. A-vit: Adaptive tokens for efficient vision transformer. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10809--10818

    2022

  40. [48]

    Yuan, Y.; Wang, Y.; Wang, L.; Zhao, X.; Lu, H.; Wang, Y.; Su, W.; and Zhang, L. 2023. Isomer: Isomerous transformer for zero-shot video object segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 966--976

    2023

  41. [49]

    Zhang, Z.; and Tao, D. 2012. Slow feature analysis for human action recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(3): 436--450

    2012

  42. [50]

    Zhu, X.; Dai, J.; Yuan, L.; and Wei, Y. 2018. Towards high performance video object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 7210--7218

    2018

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.