Pith. sign in

REVIEW 4 major objections 6 minor 43 references

Temporal Feature Distillation for Label-Efficient Precise Event Spotting in Sports Videos

T0 review · 4 major / 6 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read Aligning backbone features instead of projection heads preserves the subtle motion cues precise sports event spotting needs, so models reach high accuracy with far fewer frame labels.

desk verdict Solid label-efficient PES recipe whose core insight (backbone features, not DINO heads) is real; the 4.54-point headline is partly inflated by shared-backbone SSL re-implementations. read the letter →

arxiv 2607.10998 v1 pith:3G746N6G submitted 2026-07-13 cs.CV

Hao Xu , Xinyu Wei , Sam Wells , Sunil Aryal This is my paper · ORCID
classification cs.CV
keywords preciseeventspottingsemi-supervisedlearningsportsvideounderstandingtemporalfeaturedistillationvisiontransformersself-distillationmotion-awareaugmentation
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

Precise Event Spotting asks a model to mark the exact frame of a sports action—such as a stroke, pass, or dive entry—even when neighboring frames look almost the same. Standard image self-distillation forces representations to stay invariant across views, which washes out the tiny motion differences that mark true event boundaries. This paper argues that a different objective, Temporal Feature Distillation, keeps those boundary-sensitive cues by matching multi-scale backbone features rather than projection-head outputs, after a short supervised warm-up that first teaches the model what events look like. A gated multi-scale shift module injects temporal motion into ordinary Vision Transformers, and motion-aware augmentations force the model to focus on moving objects instead of static backgrounds. On four fine-grained sports benchmarks the method consistently beats supervised and semi-supervised baselines, matching full-supervision performance with only 80 percent of the labels on two datasets and delivering multi-point gains under 10 percent labels.

What carries the argument

Temporal Feature Distillation (TFD): a semi-supervised objective that aligns temporally structured backbone features (not projection-head outputs) between a strongly augmented student and a weakly augmented EMA teacher, so motion-sensitive and boundary-aware cues survive for frame-level localization.

What would settle it

Train the same TGS-ViT backbone with pure DINO-style projection alignment (no backbone-feature loss, no warm-up) under 10 percent labels on the four datasets; if that projection-only model matches or exceeds TFD mAP and still shows sharp inter-frame similarity drops at ground-truth event boundaries, the claim that backbone alignment is required fails.

Watch

Extended reading notes

Core claim

Direct DINO-style projection alignment is ineffective for Precise Event Spotting because it over-smooths adjacent frames and treats subtle motion as noise; aligning ℓ2-normalized backbone features from a temporally enhanced Vision Transformer, after supervised warm-up and ramp-up, preserves boundary sensitivity and yields large gains under limited frame-level labels.

Load-bearing premise

The method rests on the premise that matching backbone features after a brief supervised warm-up will keep the small motion cues that projection-head methods erase, without proving those features are intrinsically more boundary-sensitive beyond the four sports datasets tested.

Editorial extensions

If this is right

  • Semi-supervised PES can match or exceed fully supervised 100 percent baselines using only 80 percent labeled frames on some sports datasets.
  • Ordinary Vision Transformers become competitive for PES once multi-scale gated temporal shifting supplies motion information without a dedicated video architecture.
  • Motion-aware temporal augmentations that emphasize moving objects improve consistency training more than standard image augmentations in sports video settings.
  • Low-label regimes (around 10 percent annotations) benefit most, with multi-point mAP gains over prior SSL and supervised methods.
  • Pure unsupervised pretraining is insufficient when small event cues occupy only a few pixels; a supervised warm-up is required first.

Reading between the lines

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

  • The same backbone-feature alignment idea may transfer to other fine-grained temporal localization tasks outside sports, such as surgical phase detection or industrial process monitoring, where adjacent frames are visually similar.
  • If projection-level invariance systematically erases boundary cues, future video SSL methods for localization may need to default to intermediate-feature matching rather than final-head matching.
  • The need for a short supervised warm-up suggests a broader principle: when the semantic space of a short clip is narrow, pure self-distillation collapses distinctions that later task labels require.
  • Dataset-specific optimal clip length (Tennis versus figure skating) implies temporal context length should be treated as a first-class hyperparameter when porting the method to new sports or camera setups.
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 / 6 minor

Summary. The paper addresses label-efficient Precise Event Spotting (PES) in sports videos. It argues that DINO-style projection-head self-distillation is poorly suited to PES because it over-smooths adjacent-frame representations and suppresses boundary-sensitive motion cues. The authors propose Temporal Feature Distillation (TFD), which instead aligns ℓ2-normalized backbone features from a student–EMA-teacher pair (Eqs. 8–11), after a supervised warm-up and cosine ramp-up of the distillation weight. They also introduce Transformer Gate Shift (TGS), a multi-scale element-wise gated temporal shift module for ViTs (Eqs. 1–7), and Temporal Motion Augmentation (TMA) that preserves high-motion patches and applies short temporal masks. Experiments on Tennis, FSComp, FSPerf, and FineDiving under 10–80% labels report consistent gains; the headline result is +4.54 mAP over ASTRM at 10% labels on FSPerf, with 80%-label performance matching or exceeding fully supervised 100% baselines on two of four datasets (supp.).

Significance. PES is a practically important, annotation-heavy sports-video task, and semi-supervised methods for it remain underexplored relative to image SSL and coarse action recognition. The core insight—that projection-level invariance can destroy the temporal discriminability PES needs—is well motivated and directly supported by the inter-frame cosine-similarity analysis in Figure 4 and the attention maps in Figure 5. Isolating the distillation objective on a shared ViT-S+TGS backbone (Section 5.2) is a sound experimental choice for comparing learning objectives. Ablations in Table 2 cleanly separate TGS, TFD, and TMA. If the gains hold under fuller statistical reporting and clearer baseline disclosure, the work is a useful, task-aware contribution to label-efficient fine-grained video localization, with reusable modules (TGS, TMA) beyond the specific distillation recipe.

major comments (4)
  1. Table 1 / Abstract: The main quantitative claims (including the +4.54 mAP FSPerf@10% result and the 80%-vs-100% claim) are reported as point estimates only, despite the protocol of three random seeds (Section 5.2). Without standard deviations or a significance test, it is hard to judge whether several of the smaller margins (e.g., Tennis@80% 93.04 vs ASTRM 92.97; FineDiving@80% 66.57 vs ASTRM 66.19 / Kwon 67.92) are reliable. Please add mean±std (or at least std) to Table 1 and state whether the headline improvements remain significant.
  2. Section 5.2 and Table 1: All ViT-based SSL/USL baselines (DINO, Teeti et al., Kwon et al., Zhou et al.) are re-implemented on the authors’ ViT-S+TGS backbone, resolution, clip length, and schedule. This is appropriate for isolating the learning objective, but the paper never reports those same SSL objectives on their native backbones or on plain ViT without TGS. Table 2 shows that TGS alone accounts for large gains (e.g., ViT-S FSPerf@10%: 28.61 → 41.37). Please either (i) add a controlled “SSL objective on ViT-S without TGS” column, or (ii) explicitly qualify that the SSL ranking is under a shared TGS-enhanced backbone and that absolute gains relative to published native implementations may differ.
  3. Abstract / Section 5.3: The claim that “with only 80% labeled data, it matches or surpasses the fully supervised 100% baseline on two of the four datasets” cannot be verified from the main text; 100% fully supervised numbers are deferred to the supplementary material, and Table 1 stops at 80%. For a claim featured in the abstract and Figure 1, the 100% supervised reference numbers (at least for ASTRM / E2E / the authors’ own fully supervised ViT-S+TGS) should appear in the main paper, with the same evaluation protocol.
  4. Section 3.4 / Eq. (9) and the free-parameter set: The central modeling assumption—that backbone features after TGS are intrinsically more boundary-sensitive than projection-head outputs—is supported empirically by Figure 4 on one Tennis clip, but the method depends on several unablated schedule and module choices (5-epoch warm-up, 10-epoch cosine ramp of λ to 1, TGS deltas {1,3,5} and 3-way channel split, TMA 16×16 / top-9 / mask length 3, dual EMA decays). A short sensitivity study on warm-up length and λ schedule (or a statement that defaults transfer without retuning across the four datasets) is needed to show that the recipe is not brittle to these free parameters.
minor comments (6)
  1. Table 3: ViT-S+TGS costs ~1000 GFLOPs and 867 FPS vs ~40–45 GFLOPs / multi-thousand FPS for the CNN PES baselines. The paper notes that CNN/ViT families are not directly comparable, but a short discussion of when the accuracy–cost trade-off is justified for PES deployment would help readers.
  2. Table 2, clip-length ablation: L=100 is suboptimal on FSPerf (several other lengths gain +5.5 to +8.7 mAP). The choice is justified as a community default, but the main results on FSPerf may understate the method; consider reporting the best L per dataset in the supplement or noting the sensitivity more prominently.
  3. Figure 1 caption and Abstract: “matches or surpasses the fully supervised 100% baseline on 2 of 4 datasets” should name which two datasets and which 100% reference model, once those numbers are moved into the main paper.
  4. Section 3.3 / Figure 3: TMA’s motion-map construction (frame difference, top-2% pixels per patch, keep top-9 of 16×16) is clear, but the probability 0.5 and mask settings appear only in Section 5.1; a brief pointer in Section 3.3 would improve reproducibility from the method section alone.
  5. Related Work: Liu et al. [11] is correctly distinguished as multimodal; a one-sentence note on whether any pure-RGB SSL PES baseline existed before this work would sharpen the novelty claim.
  6. Typos / polish: “generalisability” (Section 2.3) is fine in British English but inconsistent with “generalization” elsewhere; “The first attempts” capitalization in Section 2.3; ensure δ vs 𝛿 notation is uniform between text and equations.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical SSL method with external-benchmark mAP claims; no tautological derivation or load-bearing self-citation chain.

full rationale

This is a standard empirical computer-vision paper proposing TGS (multi-scale gated shift for ViT), TMA, and TFD (backbone-feature cosine alignment after supervised warm-up, Eqs. 8–11). The strongest claims are measured mAP improvements on four public PES benchmarks under held-out test splits and controlled label fractions (Table 1, Abstract). The distillation objective is ordinary stop-gradient cosine distance on ℓ2-normalized features; it is not defined in terms of the reported mAP, nor is any parameter fitted to a subset and then “predicted” as the headline metric. Self-citations ([36], [37]) appear only as related-work baselines or prior PES modules and do not supply uniqueness theorems, forced ansätze, or the central performance numbers. Section 5.2’s decision to re-implement SSL baselines on a shared ViT-S+TGS backbone is a comparison-protocol choice, not a circular reduction of the result to its inputs. No self-definitional equations, fitted-input-as-prediction, or renaming of known results that collapse the claims by construction. The paper is therefore self-contained against external data; circularity score is zero.

Assumptions & free parameters 5 free parameters · 3 assumptions · 2 invented entities

The work is an empirical deep-learning method. Its claims rest on standard SSL assumptions (EMA teacher, consistency under augmentation) plus a handful of hand-chosen schedule and architecture hyper-parameters; no new physical or mathematical entities are postulated.

free parameters (5)
  • supervised warm-up length = 5 epochs
    Fixed at 5 epochs before distillation begins; chosen by authors, not derived.
  • distillation ramp-up length and max λ = 10 epochs, λ_max=1
    Cosine ramp from 0 to 1 over 10 epochs, then held at 1; schedule is hand-tuned.
  • TGS multi-scale deltas and channel split = d={1,3,5}
    Kernel sizes corresponding to d∈{1,3,5} and equal three-way channel split are design choices.
  • TMA motion-grid and mask parameters = 16×16, top-9, mask=3
    16×16 grid, top-9 cells, mask length 3, ≤5 segments, probability 0.5 are free design knobs.
  • EMA decay rates = 0.9995 / 0.99
    Base 0.9995 and fast 0.99 chosen by authors.
assumptions (3)
  • domain assumption EMA teacher provides stable targets for consistency regularization (Mean Teacher assumption).
    Invoked throughout Section 3.4; standard in SSL but not proved for PES.
  • domain assumption Frame-difference motion maps sufficiently highlight event-relevant objects without external detectors.
    Used to construct TMA (Section 3.3); may fail under heavy camera motion.
  • ad hoc to paper Backbone features after TGS retain more boundary-sensitive information than projection-head outputs.
    Central design hypothesis of TFD (Section 3.4); supported only by the empirical cosine-similarity plot.
invented entities (2)
  • Transformer Gate Shift (TGS)
    purpose: Inject multi-scale gated temporal shifts into a standard 2D ViT without a dedicated video transformer.
    New architectural module; independent evidence is limited to the ablations in Table 2.
  • Temporal Feature Distillation (TFD)
    purpose: Align temporally structured backbone features rather than DINO projection heads for PES.
    New training objective; evidence is the performance tables and Figure 4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Temporal Feature Distillation for Label-Efficient Precise Event Spotting in Sports Videos." pith.science (2026). https://pith.science/paper/3G746N6G

@misc{pith2026260710998,
  author       = {Pith},
  title        = {Pith review of: Temporal Feature Distillation for Label-Efficient Precise Event Spotting in Sports Videos},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3G746N6G}},
  note         = {Machine review of arXiv:2607.10998}
}
read the original abstract

Precise Event Spotting (PES) requires distinguishing visually similar yet semantically distinct adjacent frames, making it fundamentally different from image classification and coarse action recognition. Although self-distillation methods such as DINO have shown strong representation learning ability in images, we find that directly applying them to PES is ineffective: without supervised guidance, subtle but crucial motion cues are often suppressed as noise, leading to representations that are insensitive to precise event boundaries. To address this, we propose Temporal Feature Distillation, a semi-supervised objective that aligns temporally informative backbone features, rather than projection-head outputs, to preserve motion-sensitive and boundary-aware cues for frame-level localization. A supervised warm-up with a ramp-up schedule further stabilizes training by ensuring that meaningful event cues are learned before unlabeled distillation begins. We also introduce Transformer Gate Shift, a multi-scale gated shifting module that injects motion-aware temporal information into Vision Transformers. Experiments on four fine-grained sports benchmarks show consistent improvements over fully supervised and semi-supervised baselines. Under 10\% supervision on FSPerf, our method improves mAP by 4.54 points over the strongest competing approach, and with only 80\% labeled data, it matches or surpasses the fully supervised 100\% baseline on two of the four datasets.

Figures

Figures reproduced from arXiv: 2607.10998 by the authors.

Figure 1
Figure 1. Performance comparison across four sports datasets under varying annotation ratios. Our method consistently [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall architecture of the proposed SSL framework. The student and EMA teacher networks are built upon ViT [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Effect of the proposed Temporal Motion Augmenta [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Inter-frame cosine similarity between adjacent backbone features for TFD (ours) and DINO on a representative [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Attention map comparison between our TFD (top) [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 4 linked inside Pith

  1. [1]

    Mengqi Cao, Min Yang, Guozhen Zhang, Xiaotian Li, Yilu Wu, Gangshan Wu, and Limin Wang. 2022. SpotFormer: A transformer-based framework for precise soccer action spotting. In2022 IEEE 24th International Workshop on Multimedia Signal Processing (MMSP). IEEE, 1–6

    2022

  2. [2]

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. 2021. Emerging properties in self-supervised vision transformers. InProceedings of the IEEE/CVF international conference on computer vision. 9650–9660

    2021

  3. [3]

    Yanbei Chen, Massimiliano Mancini, Xiatian Zhu, and Zeynep Akata. 2022. Semi- supervised and unsupervised deep visual learning: A survey.IEEE transactions on pattern analysis and machine intelligence46, 3 (2022), 1327–1347

    2022

  4. [4]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929(2020)

    arXiv 2020

  5. [5]

    James Hong, Matthew Fisher, Michaël Gharbi, and Kayvon Fatahalian. 2021. Video pose distillation for few-shot, fine-grained sports action recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 9254– 9263

    2021

  6. [6]

    James Hong, Haotian Zhang, Michaël Gharbi, Matthew Fisher, and Kayvon Fa- tahalian. 2022. Spotting temporally precise, fine-grained events in video. In European Conference on Computer Vision. Springer, 33–51

    2022

  7. [7]

    Longlong Jing, Toufiq Parag, Zhe Wu, Yingli Tian, and Hongcheng Wang. 2021. Videossl: Semi-supervised learning for video classification. InProceedings of the IEEE/CVF winter conference on applications of computer vision. 1110–1119

    2021

  8. [8]

    Akash Kumar and Yogesh Singh Rawat. 2022. End-to-end semi-supervised learn- ing for video action detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 14700–14710

    2022

Show all 43 references
  1. [9]

    Donghyeon Kwon, Inho Kim, and Suha Kwak. 2025. Boosting Semi-Supervised Video Action Detection with Temporal Context. In2025 IEEE/CVF Winter Confer- ence on Applications of Computer Vision (W ACV). IEEE, 847–858

    2025

  2. [10]

    Ji Lin, Chuang Gan, and Song Han. 2019. Tsm: Temporal shift module for efficient video understanding. InProceedings of the IEEE/CVF international conference on computer vision. 7083–7093

    2019

  3. [11]

    Zhaoyu Liu, Kan Jiang, Murong Ma, Zhe Hou, Yun Lin, and Jin Song Dong. 2025. Few-Shot Precise Event Spotting via Unified Multi-Entity Graph and Distillation. arXiv preprint arXiv:2511.14186(2025)

    2025

  4. [12]

    Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101(2017)

    2017 arXiv

  5. [13]

    Sauradip Nag, Xiatian Zhu, Yi-Zhe Song, and Tao Xiang. 2022. Semi-supervised temporal action detection with proposal-free masking. InEuropean Conference on Computer Vision. Springer, 663–680

    2022

  6. [14]

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El- Nouby, et al. 2023. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193(2023)

    2023 arXiv

  7. [15]

    Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, and Piotr Dollár. 2020. Designing network design spaces. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 10428–10436

    2020

  8. [16]

    Sanchayan Santra, Vishal Chudasama, Pankaj Wasnik, and Vineeth N Balasub- ramanian. 2025. Precise Event Spotting in Sports Videos: Solving Long-Range Dependency and Class Imbalance. InProceedings of the Computer Vision and Pattern Recognition Conference. 3163–3172

    2025

  9. [17]

    Madeline C Schiappa, Yogesh S Rawat, and Mubarak Shah. 2023. Self-supervised learning for videos: A survey.Comput. Surveys55, 13s (2023), 1–37

    2023

  10. [18]

    Oriane Siméoni, Huy V Vo, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Rama- monjisoa, et al. 2025. Dinov3.arXiv preprint arXiv:2508.10104(2025)

    2025 arXiv

  11. [19]

    Ayush Singh, Aayush J Rana, Akash Kumar, Shruti Vyas, and Yogesh Singh Rawat

  12. [20]

    InProceedings of the AAAI Conference on Artificial Intelligence, Vol

    Semi-supervised active learning for video action detection. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 4891–4899

  13. [21]

    Kihyuk Sohn, David Berthelot, Nicholas Carlini, Zizhao Zhang, Han Zhang, Colin A Raffel, Ekin Dogus Cubuk, Alexey Kurakin, and Chun-Liang Li. 2020. Fixmatch: Simplifying semi-supervised learning with consistency and confidence. Advances in neural information processing systems...

    2020

  14. [22]

    Swathikiran Sudhakaran, Sergio Escalera, and Oswald Lanz. 2020. Gate-shift networks for video action recognition. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 1102–1111

    2020

  15. [23]

    Swathikiran Sudhakaran, Sergio Escalera, and Oswald Lanz. 2023. Gate-shift-fuse for video action recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 9 (2023), 10913–10928

    2023

  16. [24]

    Antti Tarvainen and Harri Valpola. 2017. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.Advances in neural information processing systems30 (2017)

    2017

  17. [25]

    Izzeddin Teeti, Rongali Sai Bhargav, Vivek Singh, Andrew Bradley, Biplab Baner- jee, and Fabio Cuzzolin. 2023. Temporal DINO: A Self-Supervised Video Strategy to Enhance Action Prediction. InProceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) Works...

    2023

  18. [26]

    Zhan Tong, Yibing Song, Jue Wang, and Limin Wang. 2022. Videomae: Masked autoencoders are data-efficient learners for self-supervised video pre-training. Advances in neural information processing systems35 (2022), 10078–10093

    2022

  19. [27]

    Kim Hoang Tran, Phuc Vuong Do, Ngoc Quoc Ly, and Ngan Le. 2024. Unifying global and local scene entities modelling for precise action spotting. In2024 International Joint Conference on Neural Networks (IJCNN). IEEE, 1–8

    2024

  20. [28]

    Jesper E Van Engelen and Holger H Hoos. 2020. A survey on semi-supervised learning.Machine learning109, 2 (2020), 373–440

    2020

  21. [29]

    Renaud Vandeghen, Anthony Cioppa, and Marc Van Droogenbroeck. 2022. Semi- supervised training to improve player and ball detection in soccer. InProceedings of the IEEE/cvf conference on computer vision and pattern recognition. 3481–3490

    2022

  22. [30]

    Xiang Wang, Shiwei Zhang, Zhiwu Qing, Yuanjie Shao, Changxin Gao, and Nong Sang. 2021. Self-supervised learning for semi-supervised temporal action proposal. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1905–1914

    2021

  23. [31]

    Artur Xarles, Sergio Escalera, Thomas B Moeslund, and Albert Clapés. 2023. Astra: An action spotting transformer for soccer videos. InProceedings of the 6th International Workshop on Multimedia Content Analysis in Sports. 93–102

    2023

  24. [32]

    Artur Xarles, Sergio Escalera, Thomas B Moeslund, and Albert Clapés. 2024. T-deed: Temporal-discriminability enhancer encoder-decoder for precise event spotting in sports videos. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 3410–3419

    2024

  25. [33]

    Kun Xia, Le Wang, Sanping Zhou, Gang Hua, and Wei Tang. 2023. Learning from noisy pseudo labels for semi-supervised temporal action localization. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 10160– 10169

    2023

  26. [34]

    Qizhe Xie, Minh-Thang Luong, Eduard Hovy, and Quoc V Le. 2020. Self- training with noisy student improves imagenet classification. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 10687–10698

    2020

  27. [35]

    Zhen Xing, Qi Dai, Han Hu, Jingjing Chen, Zuxuan Wu, and Yu-Gang Jiang

  28. [36]

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

    Svformer: Semi-supervised video transformer for action recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 18816–18826

  29. [37]

    Bo Xiong, Haoqi Fan, Kristen Grauman, and Christoph Feichtenhofer. 2021. Mul- tiview pseudo-labeling for semi-supervised learning from video. InProceedings of the IEEE/CVF international conference on computer vision. 7209–7219

    2021

  30. [38]

    Hao Xu, Arbind Agrahari Baniya, Sam Well, Mohamed Reda Bouadjenek, Richard Dazeley, and Sunil Aryal. 2025. Deep Learning for Sports Video Event Detection: Tasks, Datasets, Methods, and Challenges. arXiv:2505.03991 [cs.CV] https: //arxiv.org/abs/2505.03991

    2025

  31. [39]

    Hao Xu, Xinyu Wei, Sam Wells, and Sunil Aryal. 2025. Multi-Focus Temporal Shifting for Precise Event Spotting in Sports Videos. arXiv:2507.07381 [cs.CV] https://arxiv.org/abs/2507.07381

    2025

  32. [40]

    Jinglin Xu, Yongming Rao, Xumin Yu, Guangyi Chen, Jie Zhou, and Jiwen Lu

  33. [41]

    InProceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Finediving: A fine-grained dataset for procedure-aware action quality assessment. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2949–2958

  34. [42]

    Feixiang Zhou, Bryan Williams, and Hossein Rahmani. 2024. Towards adap- tive pseudo-label learning for semi-supervised temporal action localization. In European Conference on Computer Vision. Springer, 320–338

    2024

  35. [43]

    He Zhu, Junwei Liang, Chengzhi Lin, Jun Zhang, and Jianming Hu. 2022. A transformer-based system for action spotting in soccer videos. InProceedings of the 5th international acm workshop on multimedia content analysis in sports. 103–109

    2022

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.