Pith. sign in

REVIEW 5 major objections 5 minor 38 references

Localizing to Debias: A Patch-Level Benchmark and Baseline for Weakly Supervised Spatial Anomaly Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that weakly supervised video anomaly detection can be spatially grounded by dynamically pruning background-dominant spatio-temporal tubelets under a motion prior, and that a released patch-level benchmark with spatial…

desk verdict A genuinely useful benchmark and a sensible baseline for weakly supervised spatial anomaly detection, but the new spatial annotations are unvalidated and every spatial claim leans on them. read the letter →

arxiv 2608.12045 v1 pith:7PZRCS3L submitted 2026-08-12 cs.CV

Sara Abdulaziz , Abdulrahman Al-Abri , Giacomo D'Amicantonio , Egor Bondarev This is my paper · ORCID
classification cs.CV
keywords weaklysupervisedvideoanomalydetectionspatiallocalizationpatch-levelbenchmarkdynamictokensparsificationmotionregularizationscenebiasauditspatio-temporaltubeletsboundingboxannotations
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

This paper tries to show that the background-scene bias in weakly supervised video anomaly detection, where models are trained on video-level labels only, can be reduced by grounding predictions in space, and that spatial grounding is achievable without extra supervision. It proposes SST-WSVADL, a two-branch model whose patch branch progressively prunes background-dominant spatio-temporal tubelets under a motion prior extracted from time-reversal asymmetry, forcing the model to score dynamic, anomaly-relevant regions instead of static context. The paper also releases bounding-box annotations and a method-agnostic patch-level evaluation protocol for UCF-Crime, XD-Violence, and MSAD, so spatial localization and scene bias become measurable across methods. A sympathetic reader would care because if this holds, weakly supervised detectors can become spatially accountable without extra detectors, dense annotations, or vision-language prompts.

What carries the argument

At the center is the Dynamic Tubelet-Feature Encoder (DTFE), which converts snippet proposals into tubelets — patches stacked over time — and prunes them layer by layer with hard binary decisions. Pruning is steered by a motion score $m_i = \frac{1}{2}\|x_i - x_i^{\mathrm{rev}}\|_2$, the norm of the antisymmetric part of a tubelet's embedding under temporal reversal, so moving, anomaly-like regions survive while static background tubelets are dropped. The motion matrix $M$ enters the loss as $L_{\mathrm{motion}} = -\frac{1}{N_l}\sum_i (G_l \cdot M)$, where $G_l$ is the binary keep-decision matrix at pruning layer $l$. The retained tubelet tokens then attend to snippet-level features through Patch-Snippet Attention (PSA), and the Online Clip Proposal (OCP) module feeds the patch branch the hardest normal and abnormal clips drawn from the temporal branch's memory banks.

What would settle it

Run the proposed model on near-static anomalies such as an unattended bag or a person lying still and check whether the retained tubelets overlap the anomaly; if the motion prior prunes those tubelets, sparsification is removing the discriminative evidence it claims to preserve. Independently re-annotate a random subset of test frames and measure inter-annotator agreement with the released boxes, since low agreement would mean the spatial metrics are not anchored to a stable notion of where the anomaly is.

Watch

Extended reading notes

Core claim

The paper's central claim is that a weakly supervised spatio-temporal anomaly detector can locate where anomalies occur by progressively pruning background-dominant tubelets, with the pruning steered by a motion prior, and that this spatial grounding is compatible with competitive temporal detection. It further claims that the field's fragmented evaluation, mixing frame-level AUC, temporal IoU, mean IoU, and prompt heatmaps, hides scene bias, and that a unified patch-level benchmark with released spatial annotations, scored by temporal IoU, mean IoU, patch AUC, and patch average precision, makes such bias auditable. On its own metrics, the model improves temporal AUC and AP on MSAD by substantial margins over its temporal-only counterpart and reduces mean scene bias by a modest but consistent amount under a fixed backbone.

Load-bearing premise

The spatial boxes that define every localization and bias metric were produced by the authors' own update rules and were not checked against independent human labels, so the whole spatial evaluation inherits whatever bias those boxes carry.

Editorial extensions

If this is right

  • A weakly supervised model can localize anomalies in space without object detectors, dense spatio-temporal labels, or vision-language prompts, using video-level labels plus a motion prior.
  • The released patch-level annotations and protocol give future WSVAD methods common metrics (temporal IoU, mean IoU, patch AUC, patch AP) on three datasets, so spatial results become directly comparable.
  • Adding the sparse spatial branch with hard pruning improves temporal detection on MSAD and XD-Violence and lowers mean scene bias by 0.013 under a fixed backbone, while the backbone choice has a larger effect.
  • Hard pruning outperforms soft pruning for patch-level grounding (PAP 25.80 vs 18.69 on UCF-Crime) at a small cost in temporal AP, showing that committing tokens spatially is what sharpens localization.

Reading between the lines

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

  • This suggests a training-free variant: rank tubelets directly by their time-reversal motion score, since the score is computed before the encoder and could be used without learned pruning.
  • A natural next use of the released protocol is to score VLM- and detector-based WSVADL methods on the same PAUC/PAP axes, which the paper's comparisons could not include because their code is unavailable.
  • If the motion prior is the active debiasing ingredient, the same time-reversal regularizer could be transplanted into other video models to reduce scene-conditioned scoring outside anomaly detection.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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. This paper addresses weakly supervised video anomaly detection and localization (WSVADL). It introduces SST-WSVADL, a two-branch model with a snippet-level temporal branch and a patch-level localization branch. The patch branch uses dynamic token sparsification via a Dynamic Tubelet-Feature Encoder (DTFE) guided by a motion prior computed from time-reversed tubelet embeddings, together with a patch-snippet cross-attention module. The paper also releases per-frame spatial bounding-box annotations for UCF-Crime, XD-Violence, and MSAD, and proposes a patch-level evaluation protocol with PAUC/PAP and TIoU/MIoU metrics. Experiments report competitive temporal detection and spatial localization relative to existing methods, and a scene-bias audit using the Bias-AUC measure shows modest reductions in scene bias when the spatial branch is added. The paper is appropriately hedged, presenting the method as a transparent baseline rather than a complete solution.

Significance. If the released annotations and protocol are valid, the benchmark is a useful community resource: it provides the first unified patch-level spatial evaluation for three public WSVAD datasets and enables auditing of scene bias in anomaly predictions. The method itself is also valuable as a detector-free and non-VLM baseline that combines temporal and spatial reasoning end-to-end. Strengths include the commitment to public data release, the method-agnostic evaluation protocol, explicit reporting of limitations, and the use of an external evaluator (Bias-AUC) for the scene-bias analysis, which avoids circularity with the training objective. However, the significance is conditional on the validity of the new spatial annotations and on full disclosure of the protocol and architecture details, several of which are currently deferred to the supplementary material.

major comments (5)
  1. [Section 3 (Datasets and Spatial Annotations)] The annotation protocol is not validated. The semantically-driven update policy that keeps a single box for stationary events and updates only after substantial shifts is subjective; no inter-annotator agreement, no comparison with independent human labels, and no quantitative evidence for the claim that these boxes are superior to prior tight-box annotations are provided in the main text. The statistics are deferred to Supp. S1.1-S1.2. Because every spatial metric (TIoU, MIoU, PAUC, PAP) and the scene-bias audit are computed against these boxes, this missing validation is load-bearing for the central benchmark and localization claims. Please add an inter-annotator study or an independent label comparison and summarize the key statistics in the main text.
  2. [Section 4.2 (Patch-level WSVAD)] The core component DTFE is not defined in the main text: the full mathematical formulation is provided only in Supp. 2.1. The pruning mechanism, including retention ratios, layer-wise decision variables, and hard versus soft masks, is essential to reproduce and interpret the sparsification claims. Since the paper's stated contribution is a reproducible baseline and benchmark, this omission is load-bearing. The formulation and pruning policy should be integrated into the main text or into a self-contained appendix that is part of the reviewed manuscript.
  3. [Section 4.3 (Training Procedure), Eq. (7)] The motion loss L_motion = -1/N_l * sum(G_l · M) involves a binary keep-decision matrix G_l. It is not explained how gradients propagate through this discrete decision to update the DTFE parameters when hard pruning is used. If a straight-through estimator or a continuous relaxation is intended, it must be specified; otherwise the claimed mechanism by which motion regularization shapes learned sparsification is not supported by the equations as written.
  4. [Section 5.3 and Tables 1, 4] The spatial comparison mixes two different annotation subsets. Table 1 reports TIoUsub on the subset of frames annotated by Liu and Ma, where SST-WSVADL attains 26.25, while Table 4 reports TIoU on the full new annotations, where the VMAEv2 model attains 14.59. The text states that SST-WSVADL outperforms both the supervised approach [16] and STPrompt [33] and cites 14.59 TIoU, which is a full-set number compared against subset numbers from Table 1. This makes the claimed outperformance unverifiable as presented. The authors should report a single consistent evaluation set or clearly separate the subset and full-set comparisons.
  5. [Section 5.1 (Evaluation Metrics)] The PAUC/PAP protocol is under-specified. The description that patch scores are masked by temporal frame scores and then flattened leaves open how normal frames are handled, what threshold is applied, and how video-level imbalance enters the pooled AUC/AP computation. Because this protocol is a central contribution, a precise mathematical definition, including the treatment of negative frames, must appear in the main text rather than only in the supplementary material.
minor comments (5)
  1. [Section 5.4, Table 7] The text says that incorporating the motion loss consistently improves both temporal and spatial performance, but Table 7 shows PAP dropping from 33.23 to 25.80 when the motion loss is added. The subsequent sentence acknowledges this trade-off; please rephrase the earlier claim to be accurate at the point of first mention.
  2. [Section 5.2 (Implementation Details)] The patch size P is never specified; only the temporal split size t_z=2 is given. Please report P and the resulting spatial grid size for each dataset, as this is a free parameter that affects all spatial results.
  3. [Section 5.4, Table 5] The row labeled 'Supervised' uses ground-truth snippet labels only for proposal selection, without any spatial supervision. Calling this setting 'Supervised' is misleading; please rename it to something like 'GT-proposal sampling'.
  4. [Introduction / Data Availability] The manuscript mentions 'GitHub Code' but provides no URL or explicit data availability statement. For a benchmark contribution, the final version should include the repository link and the exact license and access conditions for the released annotations.
  5. [Section 4.1 (Snippet-level WSVAD)] The sentence 'the exact temporal evidence provided by the MIL loss in the temporal branch is applied to the patch branch' is unclear. Equation (8) writes L_P = L_WSVAD + lambda_5*L_motion, but it is not stated whether the same MIL snippet pairs and labels are reused in the patch branch or how the two branches exchange the evidence. Please clarify the shared supervision.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the model is trained with losses independent of the evaluation metrics, and the released spatial annotations are used only for evaluation, not for training.

full rationale

The paper's central derivations are not circular. SST-WSVADL is trained end-to-end with a binary cross-entropy loss plus auxiliary losses (Eq. 3, Eq. 8), and the evaluation metrics (AUC, AP, TIoU, MIoU, PAUC, PAP) never appear as training objectives, so there is no fitted-input-called-prediction pattern. The spatial branch receives no ground-truth box supervision: the Online Clip Proposal module selects snippets using temporal prototypes, the patch branch converts these into tubelets, and the patch objective is the same MIL-style BCE with a motion-regularization term computed from time-reversal feature norms. The newly released bounding-box annotations are used only for evaluation and benchmarking, not as training labels, so the localization benchmark is not self-defined relative to the model's optimization. The scene-bias analysis adopts the Bias-AUC audit from the authors' prior work [1]; this is a self-citation, but it is an external evaluator applied post hoc on normal frames, it is not a term in the loss, and it does not invoke a uniqueness theorem or forbid alternative metrics. The citation provides an independently applicable measure, so per the review rules it does not raise the circularity score. The paper's ablations (Tables 5-10) test components against external and newly introduced metrics without reusing those metrics as losses. Remaining concerns, such as the absence of inter-annotator agreement for the new spatial annotations, are validation and robustness issues rather than evidence of circular derivation.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

SST-WSVADL introduces new model components (DTFE, PSA, OCP) but these are architectural modules, not postulates of new physical entities, so the invented-entities list is empty. The axioms above are the domain assumptions that the empirical claims rely on; none are standard formal axioms.

free parameters (4)
  • motion loss weight lambda_5 = 0.01
    Chosen by ablation on UCF-Crime (Table 7); sensitivity shown across 0.01, 0.001, and 0.0001.
  • temporal split size t_z = 2
    Set empirically for all datasets (Section 5.2); controls tubelet temporal extent.
  • top-k proposal selection = k unspecified; rank=1 selected
    OCP picks the first of the top-k normal/abnormal features; Table 5 justifies rank 1, but k is not stated.
  • patch size P = not specified
    Spatial grid is H/P x W/P but P is not given in the main text; needed to reproduce tubelet generation.
assumptions (4)
  • domain assumption Time-reversal antisymmetry isolates motion cues in tubelet embeddings
    Eqs. (5)-(6) set motion score to the norm of (x - x_rev)/2, assuming appearance is time-symmetric and motion is not; lighting transients, occlusions, or camera effects also break time symmetry, so the score can conflate appearance and motion. Entered in Section 4.3.
  • domain assumption The temporal branch's memory-bank top-k selections identify hard discriminative snippets
    OCP (Section 4.1) assumes UR-DMU's prototype similarity ranking reliably proposes normal and abnormal clips for patch training; no independent verification that the chosen clips contain localized anomaly evidence.
  • domain assumption A single semantic bounding box per event is a valid spatial ground truth
    Section 3's update policy keeps one box per stationary event and updates only after substantial shifts, prioritizing semantic continuity over frame precision; this can under-localize dispersed multi-actor events and is not validated by inter-annotator agreement.
  • domain assumption Masking patch scores by temporal frame scores yields a meaningful patch-level metric
    Section 5.1 defines PAUC/PAP after masking; if temporal scores are inaccurate or the rule is arbitrary, the patch metrics can be distorted, and no sensitivity analysis of the masking step is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Localizing to Debias: A Patch-Level Benchmark and Baseline for Weakly Supervised Spatial Anomaly Detection." pith.science (2026). https://pith.science/paper/7PZRCS3L

@misc{pith2026260812045,
  author       = {Pith},
  title        = {Pith review of: Localizing to Debias: A Patch-Level Benchmark and Baseline for Weakly Supervised Spatial Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7PZRCS3L}},
  note         = {Machine review of arXiv:2608.12045}
}
read the original abstract

Despite growing interest in weakly supervised video anomaly detection (WSVAD), current methods struggle to bridge the gap between coarse temporal supervision and fine-grained spatial reasoning. A key obstacle is the tendency of temporal detectors to latch onto background and scene-level cues rather than truly discriminative anomaly evidence. This background bias raises ethical concerns: models may inadvertently associate anomalies with societal or environmental context rather than authentic crime-related cues. Without spatial grounding, such biases remain hidden and unauditable. To address this, we propose SST-WSVADL, a sparse spatio-temporal framework that bridges temporal anomaly detection with fine-grained spatial localization. Rather than processing all spatial regions indiscriminately, SST-WSVADL progressively focuses on the most anomaly-relevant spatio-temporal regions through dynamic sparsification, naturally suppressing background dominant content while preserving discriminative evidence. The temporal and spatial branches are coupled end-to-end via motion-aware regularization that guides sparsification toward dynamically informative regions, without relying on external detectors or vision-language prompts. We publicly release frame-level spatial annotations and a method-agnostic evaluation protocol for three public datasets: UCF-Crime, XD-Violence, and MSAD. These resources enable the community to audit spatial biases in WSVAD predictions, supporting progress toward more ethical and accountable anomaly detection. Experiments demonstrate that SST-WSVADL is competitive with prior methods across benchmarks while enabling localization and patch-level auditability of scene bias, providing a reproducible foundation for interpretability-oriented evaluation of WSVAD models.

Figures

Figures reproduced from arXiv: 2608.12045 by the authors.

Figure 1
Figure 1. Overview of the proposed SST-WSVADL. The framework consists of two inter￾acting branches. (Left) The Snippet-WSVAD branch is a standard WSVAD model [37] that processes snippet-level features extracted via a pre-trained task-agnostic back￾bone [27]. The normal and abnormal clips for the patch branch are selected through the Online Clip Proposal (OCP) module. (Right) The Patch-WSVAL branch converts snippet proposals i… view at source ↗
Figure 2
Figure 2. Qualitative results (left to right) on UCF-Crime, MSAD, and XD-Violence. Ex￾amples are from Arrest024, Arson018, Explosion022, Assault010 (UCF-Crime); Fire-24, TrafficAccident-2, Assault-11, Shooting-4 (MSAD); and v=0qtIjyt-7wg, v=DD3jfKr8e￾k, Black.Hawk.Down.2001 (XD-Violence). Zoom in for better visualization. the amount of retained tokens according to the learned pruning rates (Table S7). In UCF-Crime, roughly 25… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 25 canonical work pages

  1. [16]

    In: Proceedings of the 27th ACM International Conference on Multimedia

    Liu, K., Ma, H.: Exploring background-bias for anomaly detection in surveillance videos. In: Proceedings of the 27th ACM International Conference on Multimedia. pp. 1490–1499 (2019) 16 S. Abdulaziz et al

    work page 2019

  2. [33]

    A Benchmark and Baseline for WSVAD Localization 17 In: Proceedings of the 32nd ACM International Conference on Multimedia

    Wu, P., Zhou, X., Pang, G., Yang, Z., Yan, Q., Wang, P., Zhang, Y.: Weakly supervisedvideoanomalydetectionandlocalizationwithspatio-temporalprompts. A Benchmark and Baseline for WSVAD Localization 17 In: Proceedings of the 32nd ACM International Conference on Multimedia. pp. 9301–9310 (2024)

    work page 2024

  3. [1]

    In: Proceedings of the ECCV 2026 Workshop on Empirical Theory (2026)

    Abdulaziz, S., Bondarev, E.: Auditing frame-level AUC in weakly supervised video anomaly detection: Granularity, resolution, and scene bias. In: Proceedings of the ECCV 2026 Workshop on Empirical Theory (2026)

    work page 2026

  4. [2]

    In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Carreira, J., Zisserman, A.: Quo vadis, action recognition? a new model and the kinetics dataset. In: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 6299–6308 (2017)

    2017

  5. [3]

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

    Chen, W., Ma, K.T., Yew, Z.J., Hur, M., Khoo, D.A.A.: Tevad: Improved video anomaly detection with captions. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5549–5559 (2023)

    work page 2023

  6. [4]

    In: Proceedings of the AAAI conference on artificial intelligence

    Chen, Y., Liu, Z., Zhang, B., Fok, W., Qi, X., Wu, Y.C.: Mgfn: Magnitude- contrastive glance-and-focus network for weakly-supervised video anomaly detec- tion. In: Proceedings of the AAAI conference on artificial intelligence. vol. 37, pp. 387–395 (2023)

    2023

  7. [5]

    In: CVPR 2011

    Cui, X., Liu, Q., Gao, M., Metaxas, D.N.: Abnormal detection using interaction energy potentials. In: CVPR 2011. pp. 3161–3167. IEEE (2011)

    work page 2011

  8. [6]

    Hara, K., Kataoka, H., Satoh, Y.: Can spatiotemporal 3d cnns retrace the history of 2d cnns and imagenet? In: Proceedings of the IEEE conference on Computer Vision and Pattern Recognition. pp. 6546–6555 (2018)

    work page 2018

Show all 38 references
  1. [7]

    In: 2009 IEEE 12th International Conference on Computer Vision

    Hospedales, T., Gong, S., Xiang, T.: A markov clustering topic model for mining behaviour in video. In: 2009 IEEE 12th International Conference on Computer Vision. pp. 1165–1172. IEEE (2009)

    2009

  2. [8]

    In: European conference on computer vision

    Kong, Z., Dong, P., Ma, X., Meng, X., Niu, W., Sun, M., Shen, X., Yuan, G., Ren, B.,Tang,H.,etal.:Spvit:Enablingfastervisiontransformersvialatency-awaresoft token pruning. In: European conference on computer vision. pp. 620–640. Springer (2022)

    2022

  3. [9]

    In: 2009 IEEE conference on computer vision and pattern recognition

    Kratz, L., Nishino, K.: Anomaly detection in extremely crowded scenes using spatio-temporal motion pattern models. In: 2009 IEEE conference on computer vision and pattern recognition. pp. 1446–1453. IEEE (2009)

    2009

  4. [10]

    arXiv preprint arXiv:1901.10364 (2019)

    Landi, F., Snoek, C.G., Cucchiara, R.: Anomaly locality in video surveillance. arXiv preprint arXiv:1901.10364 (2019)

    2019 arXiv

  5. [11]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Lee, J., Bang, C., Kim, M., Cho, M.: Refinevad: Semantic-guided feature recalibra- tion for weakly supervised video anomaly detection. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 40, pp. 5899–5907 (2026)

    2026

  6. [12]

    In: European Conference on Computer Vision

    Li, G., Cai, G., Zeng, X., Zhao, R.: Scale-aware spatio-temporal relation learning for video anomaly detection. In: European Conference on Computer Vision. pp. 333–350. Springer (2022)

    2022

  7. [13]

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

    Li, M., Xu, X., Fan, H., Zhou, P., Liu, J., Liu, J.W., Li, J., Keppo, J., Shou, M.Z., Yan, S.: Stprivacy: Spatio-temporal privacy-preserving action recognition. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 5106–5115 (2023)

    2023

  8. [14]

    IEEE transactions on pattern analysis and machine intelligence 36(1), 18–32 (2013)

    Li, W., Mahadevan, V., Vasconcelos, N.: Anomaly detection and localization in crowded scenes. IEEE transactions on pattern analysis and machine intelligence 36(1), 18–32 (2013)

    2013

  9. [15]

    In: Proceedings of the AAAI conference on artificial intelligence

    Liu, K., Liu, W., Gan, C., Tan, M., Ma, H.: T-c3d: Temporal convolutional 3d network for real-time action recognition. In: Proceedings of the AAAI conference on artificial intelligence. vol. 32 (2018)

    2018

  10. [17]

    In: Pro- ceedings of the IEEE international conference on computer vision

    Lu, C., Shi, J., Jia, J.: Abnormal event detection at 150 fps in matlab. In: Pro- ceedings of the IEEE international conference on computer vision. pp. 2720–2727 (2013)

    2013

  11. [18]

    In: Proceedings of the Computer Vi- sion and Pattern Recognition Conference

    Majhi, S., D’Amicantonio, G., Dantcheva, A., Kong, Q., Garattoni, L., Francesca, G., Bondarev, E., Brémond, F.: Just dance with pi! a poly-modal inductor for weakly-supervised video anomaly detection. In: Proceedings of the Computer Vi- sion and Pattern Recognition Conference....

    2025

  12. [19]

    Mumford, D., Fogarty, J., Kirwan, F.: Geometric invariant theory, vol. 34. Springer Science & Business Media (1994)

    1994

  13. [20]

    Advances in neural infor- mation processing systems34, 13937–13949 (2021)

    Rao, Y., Zhao, W., Liu, B., Lu, J., Zhou, J., Hsieh, C.J.: Dynamicvit: Efficient vision transformers with dynamic token sparsification. Advances in neural infor- mation processing systems34, 13937–13949 (2021)

    2021

  14. [21]

    Advances in neural information processing systems27(2014)

    Simonyan, K., Zisserman, A.: Two-stream convolutional networks for action recog- nition in videos. Advances in neural information processing systems27(2014)

    2014

  15. [22]

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

    Sultani, W., Chen, C., Shah, M.: Real-world anomaly detection in surveillance videos. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 6479–6488 (2018)

    2018

  16. [23]

    In: 2023 IEEE International Conference on Multimedia and Expo (ICME)

    Sun, S., Gong, X.: Long-short temporal co-teaching for weakly supervised video anomaly detection. In: 2023 IEEE International Conference on Multimedia and Expo (ICME). pp. 2711–2716. IEEE (2023)

    2023

  17. [24]

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

    Tian, Y., Pang, G., Chen, Y., Singh, R., Verjans, J.W., Carneiro, G.: Weakly- supervised video anomaly detection with robust temporal feature magnitude learn- ing. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 4975–4986 (2021)

    2021

  18. [25]

    In: Proceedings of the IEEE inter- national conference on computer vision

    Tran, D., Bourdev, L., Fergus, R., Torresani, L., Paluri, M.: Learning spatiotem- poral features with 3d convolutional networks. In: Proceedings of the IEEE inter- national conference on computer vision. pp. 4489–4497 (2015)

    2015

  19. [26]

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

    Wang, J., Jiao, J., Bao, L., He, S., Liu, Y., Liu, W.: Self-supervised spatio-temporal representation learning for videos by predicting motion and appearance statis- tics. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4006–4015 (2019)

    2019

  20. [27]

    Wang, L., Huang, B., Zhao, Z., Tong, Z., He, Y., Wang, Y., Wang, Y., Qiao, Y.: Videomae v2: Scaling video masked autoencoders with dual masking (2023)

    2023

  21. [28]

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

    Wang, L., Li, W., Li, W., Van Gool, L.: Appearance-and-relation networks for video classification. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1430–1439 (2018)

    2018

  22. [29]

    In: Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition

    Wang, L., Tong, Z., Ji, B., Wu, G.: Tdn: Temporal difference networks for efficient action recognition. In: Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition. pp. 1895–1904 (2021)

    2021

  23. [30]

    In: European conference on computer vision

    Wang, L., Xiong, Y., Wang, Z., Qiao, Y., Lin, D., Tang, X., Van Gool, L.: Tem- poral segment networks: Towards good practices for deep action recognition. In: European conference on computer vision. pp. 20–36. Springer (2016)

    2016

  24. [31]

    In: Pro- ceedings of the IEEE conference on computer vision and pattern recognition

    Wang, X., Girshick, R., Gupta, A., He, K.: Non-local neural networks. In: Pro- ceedings of the IEEE conference on computer vision and pattern recognition. pp. 7794–7803 (2018)

    2018

  25. [32]

    arXiv preprint arXiv:2108.03825 (2021)

    Wu, J., Zhang, W., Li, G., Wu, W., Tan, X., Li, Y., Ding, E., Lin, L.: Weakly- supervised spatio-temporal anomaly detection in surveillance video. arXiv preprint arXiv:2108.03825 (2021)

    2021 arXiv

  26. [34]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Wu, P., Zhou, X., Pang, G., Zhou, L., Yan, Q., Wang, P., Zhang, Y.: Vadclip: Adapting vision-language models for weakly supervised video anomaly detection. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 6074–6082 (2024)

    2024

  27. [35]

    In: 2025 IEEE/CVF Winter Conference on Appli- cations of Computer Vision (WACV)

    Yun, S., Masukawa, R., Na, M., Imani, M.: Missiongnn: Hierarchical multimodal gnn-based weakly supervised video anomaly recognition with mission-specific knowledge graph generation. In: 2025 IEEE/CVF Winter Conference on Appli- cations of Computer Vision (WACV). pp. 4736–4745....

    2025

  28. [36]

    In: CVPR 2011

    Zhao, B., Fei-Fei, L., Xing, E.P.: Online detection of unusual events in videos via dynamic sparse coding. In: CVPR 2011. pp. 3313–3320. IEEE (2011)

    2011

  29. [37]

    In: Proceedings of the AAAI Confer- ence on Artificial Intelligence

    Zhou, H., Yu, J., Yang, W.: Dual memory units with uncertainty regulation for weakly supervised video anomaly detection. In: Proceedings of the AAAI Confer- ence on Artificial Intelligence. vol. 37, pp. 3769–3777 (2023)

    2023

  30. [38]

    IEEE Journal of Selected Topics in Signal Pro- cessing7(1), 91–101 (2012)

    Zhu, Y., Nayak, N.M., Roy-Chowdhury, A.K.: Context-aware activity recognition and anomaly detection in video. IEEE Journal of Selected Topics in Signal Pro- cessing7(1), 91–101 (2012)

    2012

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.