Pith. sign in

REVIEW 4 major objections 5 minor 46 references

Towards Scalable Video Anomaly Retrieval: A Synthetic Video-Text Benchmark

T0 review · 4 major / 5 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 SVTA, a synthetic text-video benchmark of 41,315 pairs over 68 anomaly and 30 normal classes, claims generative data can replace scarce real-world footage for training cross-modal anomaly retrieval models that transfer zero-shot to real…

desk verdict SVTA is a genuinely new large-scale synthetic anomaly-retrieval dataset, but the zero-shot transfer numbers are compromised because the test captions overlap the training captions. read the letter →

arxiv 2506.01466 v1 pith:MM2MHL5R submitted 2025-06-02 cs.CV

Shuyu Yang , Yilun Wang , Yaxiong Wang , Li Zhu , Zhedong Zheng This is my paper
classification cs.CV
keywords videoanomalyretrievalsyntheticvideo-textbenchmarktext-to-videogenerationzero-shotcross-modallong-tailcategoriesprivacy-preservingdatasetevaluation
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 SVTA, which it calls the first large-scale dataset for cross-modal video anomaly retrieval: 41,315 video-text pairs and 1.36 million frames spanning 68 anomalous and 30 normal event categories. Because real anomalous footage is scarce, long-tailed, and privacy-sensitive, the authors generate both the captions and the videos: off-the-shelf LLMs produce diverse event descriptions, and the text-to-video model Wan2.1 renders matching clips. The central claim is that this synthetic corpus can support cross-modal anomaly retrieval at scale and that models trained on it transfer to real-world anomaly benchmarks. Evaluations with CLIP4Clip, X-CLIP, and GRAM show the dataset is challenging, and zero-shot evaluation on UCFCrime-AR and OOPS! gives competitive retrieval results, supporting synthetic data as a viable substitute for real anomaly footage.

What carries the argument

The load-bearing object is the generative pipeline: LLMs turn 23 seed anomaly categories into 50 diverse captions each, and Wan2.1, a 14B text-to-video diffusion model, renders each caption into a 33-frame, 15 fps, 480x832 clip that becomes the video half of a paired sample. Category labels for samples without source annotations are assigned by an LLM, unified by K-means clustering, and manually refined, yielding the final 30-normal and 68-anomaly taxonomy organized into four anomaly groups. The retrieval baselines CLIP4Clip, X-CLIP, and GRAM then test whether video-text alignment learned from these generated pairs generalizes to real footage.

What would settle it

Take a random sample of, say, 200 SVTA clips, show each to independent annotators with its paired caption, and ask whether the video visibly depicts the described event; if more than about 10% are judged not to depict it, the zero-shot transfer results cannot be taken as evidence of visual anomaly understanding. A corroborating check is to compare against a caption-only retriever that matches query text to stored captions and ignores video content, since similar scores would indicate language priors rather than video grounding.

Watch

Extended reading notes

Core claim

The central claim is that a large-scale cross-modal anomaly retrieval benchmark can be built entirely from generated content. SVTA comprises 41,315 video-text pairs covering 68 anomaly types and 30 normal activities, produced by expanding seed anomaly categories with LLM-written descriptions and rendering them with the text-to-video model Wan2.1. The authors report that models trained on SVTA achieve competitive zero-shot retrieval on two real-world benchmarks, with GRAM reaching R@1 of 34.5% on UCFCrime-AR and X-CLIP (ViT-B/16) reaching R@1 of 18.9% on the filtered OOPS! set, alongside a best SVTA R@1 of 57.3%. The intended upshot is that synthetic footage can supply the scale, category diversity, and privacy compliance that real anomaly collection cannot.

Load-bearing premise

The benchmark rests on the assumption that Wan2.1 generates videos that faithfully show the events described in their captions; the only validation reported is informal human inspection, so if the generated clips do not actually depict the described anomalies, SVTA may measure text-to-text similarity rather than video understanding of the event.

Editorial extensions

If this is right

  • If the zero-shot transfer holds up, synthetic footage can stand in for real CCTV footage when training anomaly retrieval, removing the privacy barrier that has capped real anomaly datasets at roughly 1,900 clips.
  • A model trained on SVTA can be evaluated over 68 anomaly types rather than the 13 in UCFCrime-AR, enabling finer-grained and more realistic retrieval evaluation.
  • Because SVTA is generated, its category space can be extended on demand by prompting the LLM and generator again, making anomaly retrieval benchmarks scalable to novel events.
  • The 2:3 normal-to-anomaly ratio plus explicit category labels supports training and evaluating models for both retrieval and downstream anomaly analysis beyond the binary normal and abnormal framing.

Reading between the lines

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

  • Whether SVTA teaches visual grounding is an open question: the paper validates generator fidelity only by informal human inspection, so a systematic alignment audit would tell whether the zero-shot transfer comes from video understanding or from caption-text priors.
  • Because SVTA's OOPS! evaluation uses a filtered subset with one caption per video, a natural next check is to run the same zero-shot models on the unfiltered OOPS! benchmark to see how much the filtering affects the reported numbers.
  • A diagnostic the paper does not report is per-category retrieval accuracy across the four anomaly taxonomies, which would show whether models fail on environmental anomalies such as explosions and natural disasters or on social violations such as fighting and theft.
  • The same generative pipeline could become an augmentation strategy for real anomaly datasets, mixing synthetic clips with real footage during training rather than using synthetic data only as a standalone benchmark.
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 / 5 minor

Summary. This paper introduces SVTA, a synthetic video-text benchmark for cross-modal video anomaly retrieval. The dataset contains 41,315 video-text pairs (33 frames, 15 fps) generated by using LLM-written captions from UCA, UCFCrime-AR, OOPS!, and LLM-expanded anomaly categories to prompt the Wan2.1 text-to-video model. The authors benchmark CLIP4Clip, X-CLIP, and GRAM on SVTA and report zero-shot transfer to UCFCrime-AR and OOPS!. The main contributions claimed are the dataset's scale, category coverage (68 anomaly and 30 normal types), diversity, and privacy-safe construction.

Significance. The dataset is potentially useful as a large-scale training resource for text-based anomaly retrieval, and the reproducible generation pipeline (fixed random seed, open-source models) is a strength. However, the paper's main quantitative claim—that synthetic training transfers to real-world benchmarks—is not established as reported, because the zero-shot evaluation captions are drawn from the same text sources used to construct the training set, and because the paper provides no quantitative evidence that the synthetic videos depict the events in their captions. The baseline benchmark on SVTA itself is informative, but the validity of SVTA as a measure of visual grounding depends on controls that are currently missing.

major comments (4)
  1. [3.1, 4.2, Tables 3-4] The zero-shot evaluation protocol has a text-side leakage problem. Section 3.1 states that SVTA text descriptions are obtained by retaining all video descriptions from UCA and UCFCrime-AR and filtering OOPS! descriptions. Tables 3 and 4 then evaluate zero-shot retrieval on UCFCrime-AR and OOPS! using those same datasets as test targets. Since the exact or near-identical test captions are present in the SVTA training set, a model can rank ground-truth videos highly by text-to-text similarity without learning any visual-semantic understanding from synthetic videos. The reported gains (e.g., R@1 = 34.5 in Table 3) cannot therefore be attributed to synthetic-to-real visual transfer. A valid protocol would remove overlapping captions from training, or use novel query paraphrases for the test queries, and report both settings.
  2. [3.2] The core assumption that Wan2.1-generated videos faithfully depict the paired captions is validated only by the statement 'human inspection reveals that the content of generated videos is realistic and reasonable to some extent' (Section 3.2). No quantitative alignment measure, sample size, or inter-annotator agreement is provided. If the videos frequently fail to depict the described anomaly, SVTA becomes a text-to-text benchmark, and the cross-domain results in Tables 3 and 4 are not evidence of video understanding. The paper should include a quantitative video-text alignment evaluation (e.g., human verification on a random sample, CLIP-style similarity, or generation-based metrics) and report per-category agreement.
  3. [3.3] The claim that LLM-based attribute annotation 'exceeds 97%' accuracy is not verifiable as reported. The text says 'We manually evaluate 10% of the annotated results through random sampling' but does not specify the sample size, the number of annotators, the annotation instructions, or how disagreements were resolved. Please provide a precise evaluation protocol with confidence intervals.
  4. [4.1, Table 4] The OOPS! zero-shot evaluation uses a modified test set: 'our test set differs from the original OOPS! benchmark. We exclude videos lacking complete anomaly descriptions and retain only one matched text per video, resulting in 3,468 curated video-text pairs.' The paper does not report the overlap between these 3,468 captions and the text sources used to build SVTA (Section 3.1), nor does it compare results on the official OOPS! split. Without this, it is unclear whether the reported R@1 values reflect transfer or text memorization, and the results are not directly comparable to prior work.
minor comments (5)
  1. [Figure 3(a)] The percentages in Figure 3(a) appear to be swapped; the text states 25,170 anomaly and 16,145 normal videos, which corresponds to approximately 60.9% anomaly and 39.1% normal, the opposite of the labels shown in the figure.
  2. [Section 4.2] The sentence 'X-CLIP [22] initialized with CLIP (ViT-B/32)' cites CLIP4Clip's reference number; it should cite X-CLIP [24].
  3. [References] Reference [13] contains a typo: 'PCVPR' should be 'CVPR'.
  4. [Section 3.1] The paper does not report the number of descriptions contributed by each source (UCA, UCFCrime-AR, OOPS!, and LLM-expanded); please add a breakdown to clarify how the total of 41,315 is composed.
  5. [Table 1] Table 1 lists #Texts as 23,542 for UCA and 1,900 for UCFCrime-AR; the relationship between these counts and the SVTA caption pool should be stated explicitly.

Circularity Check

1 steps flagged · score 6.0 of 10

Zero-shot transfer numbers are inflated by text-source overlap: SVTA training captions are drawn from UCFCrime-AR and OOPS!, the same datasets used for zero-shot evaluation.

  1. fitted input called prediction [Section 3.1 (Diverse Video Descriptions Acquisition) and Section 4.2 (Cross-domain generalization, Tables 3-4)]
    "We adopt UCA [44], UCFCrime-AR [41], and OOPS! [8] as text sources due to their rich diversity of normal and anomalous video descriptions. We retain all video descriptions from UCA and UCFCrime-AR, while filtering ambiguous texts from OOPS! [...] we conduct domain generalization experiments on two real-world datasets: UCFCrime-AR [41] and OOPS! [8]. Models trained on SVTA are directly evaluated on these datasets without additional training or fine-tuning."

    The zero-shot transfer evaluation is not independent of the training input: SVTA's text set is constructed by retaining all UCFCrime-AR descriptions and filtering OOPS! descriptions, and the same two datasets are then used as the test targets. The models' text encoders are therefore optimized on captions drawn from the evaluation benchmarks, so the reported R@1 values on UCFCrime-AR and OOPS! (e.g., GRAM 34.5% T2V) can be inflated by text-side familiarity with the source annotation style and content, independent of whether the synthetic videos faithfully depict the anomalies. The conclusion that 'models trained on SVTA achieve competitive zero-shot retrieval performance' is thus partially forced by construction rather than demonstrated as synthetic-to-real visual generalization.

full rationale

The paper's dataset construction is not circular: the benchmark is a new artifact, and its internal SVTA retrieval results in Table 2 are self-contained evidence that the dataset is challenging for existing video-text retrieval methods. The circularity is located in the cross-domain generalization claim in Section 4.2. Because Section 3.1 explicitly names UCFCrime-AR and OOPS! as text sources for SVTA, and Section 4.2 evaluates zero-shot retrieval on those same datasets, the text modality is not truly zero-shot: the text encoder has been trained on captions sampled from the evaluation benchmarks. This does not fully erase the contribution, since the video side is synthetic and the test videos are real, but the central transfer result cannot be attributed cleanly to visual understanding or domain transfer. The paper's other weak point, human-inspection-only validation of text-to-video alignment, is a correctness and robustness concern rather than circularity, because it is not a derivation step that reduces to an input.

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

The central claims rely on the fidelity of the generative pipeline (text-to-video alignment) and on the representativeness of the caption sources and manual annotations. No numeric parameters are fitted to data in the analysis; generation settings such as quality=6 and TeaCache threshold=0.08 are reported but are not fitted parameters. No new theoretical entities are introduced.

assumptions (4)
  • domain assumption The text-to-video model Wan2.1 produces videos whose visual content matches the input text descriptions.
    This is the core alignment premise for the whole dataset. Section 3.2 only says 'human inspection reveals that the content of generated videos is realistic and reasonable to some extent', with no quantitative verification.
  • domain assumption The LLM-generated caption expansions preserve the semantics of the 23 base anomaly categories.
    Section 3.1 states that LLMs expand 23 categories into 50 descriptions each and that manual filtering retains 50 per category, but no automated fidelity check is described.
  • domain assumption Manual filtering and K-means refinement yield accurate 30-normal and 68-anomaly category labels.
    Section 3.3 reports self-evaluated accuracy 'exceeds 97%' on a 10% random sample, but the evaluation protocol is not described and no inter-annotator agreement is provided.
  • domain assumption The text sources UCA, UCFCrime-AR and OOPS! are representative of real-world anomaly queries.
    These datasets provide both the training captions and, later, the zero-shot test sets, as described in Sections 3.1 and 4.2. Their representativeness is assumed, not tested.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Scalable Video Anomaly Retrieval: A Synthetic Video-Text Benchmark." pith.science (2026). https://pith.science/paper/MM2MHL5R

@misc{pith2026250601466,
  author       = {Pith},
  title        = {Pith review of: Towards Scalable Video Anomaly Retrieval: A Synthetic Video-Text Benchmark},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MM2MHL5R}},
  note         = {Machine review of arXiv:2506.01466}
}
read the original abstract

Video anomaly retrieval aims to localize anomalous events in videos using natural language queries to facilitate public safety. However, existing datasets suffer from severe limitations: (1) data scarcity due to the long-tail nature of real-world anomalies, and (2) privacy constraints that impede large-scale collection. To address the aforementioned issues in one go, we introduce SVTA (Synthetic Video-Text Anomaly benchmark), the first large-scale dataset for cross-modal anomaly retrieval, leveraging generative models to overcome data availability challenges. Specifically, we collect and generate video descriptions via the off-the-shelf LLM (Large Language Model) covering 68 anomaly categories, e.g., throwing, stealing, and shooting. These descriptions encompass common long-tail events. We adopt these texts to guide the video generative model to produce diverse and high-quality videos. Finally, our SVTA involves 41,315 videos (1.36M frames) with paired captions, covering 30 normal activities, e.g., standing, walking, and sports, and 68 anomalous events, e.g., falling, fighting, theft, explosions, and natural disasters. We adopt three widely-used video-text retrieval baselines to comprehensively test our SVTA, revealing SVTA's challenging nature and its effectiveness in evaluating a robust cross-modal retrieval method. SVTA eliminates privacy risks associated with real-world anomaly collection while maintaining realistic scenarios. The dataset demo is available at: [https://svta-mm.github.io/SVTA.github.io/].

Figures

Figures reproduced from arXiv: 2506.01466 by the authors.

Figure 1
Figure 1. Due to the long-tail distribution of real-world anomaly and privacy constraints, large-scale data collection remains chal￾lenging (left). To address this limitation, we construct a large-scale Synthetic Video-Text Anomaly (SVTA) benchmark (right), which comprises 41,315 video-text pairs covering 68 anomaly types and 30 normal events. Abstract Video anomaly retrieval aims to localize anomalous events in videos using … view at source ↗
Figure 2
Figure 2. Pipeline of our Synthetic Video-Text Anomaly (SVTA) benchmark construction. First, we collect and generate diverse video descriptions via LLM. Second, we leverage a state-of-the-art open-source video generative model to craft high-quality videos. Third, we adopt LLM to assign preliminary attributes for samples lacking explicit normal/anomaly labels and refine all labels by K-Means clustering and manual verification.… view at source ↗
Figure 3
Figure 3. Dataset Statistics. An overview of the SVTA attribute annotations, including the distribution of (a) normal and anomaly videos, (b) normal categories, and (c) anomaly categories. 2410 2452 2551 2560 2580 2614 2759 3731 3960 3974 4109 4547 4824 5023 13041 0 5000 10000 15000 20000 blue people two screen fell trying down side walked road woman white black car man (a) Distributions of word frequency. 813 881 1749 2335 1… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Text statistics: (a) statistical histogram distributions of word frequency on SVTA, and (b) the sentence length in the video caption of the SVTA dataset. falling, fighting, sleeping, crawling, having a seizure, lay￾ing down, dancing, stealing, rotating 360 degrees, shu…
Figure 5
Figure 5. Figure 5: Some retrieved examples of GRAM on SVTA. We visualize top 3 retrieved videos ( [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages

  1. [1]

    Ub- normal: New benchmark for supervised open-set video anomaly detection

    Andra Acsintoae, Andrei Florescu, Mariana-Iuliana Georgescu, Tudor Mare, Paul Sumedrea, Radu Tudor Ionescu, Fahad Shahbaz Khan, and Mubarak Shah. Ub- normal: New benchmark for supervised open-set video anomaly detection. In CVPR, pages 20143–20153, 2022. 2, 3, 5

    work page 2022

  2. [2]

    Robust real-time unusual event detection using mul- tiple fixed-location monitors

    Amit Adam, Ehud Rivlin, Ilan Shimshoni, and Daviv Reinitz. Robust real-time unusual event detection using mul- tiple fixed-location monitors. IEEE transactions on pattern analysis and machine intelligence, 30(3):555–560, 2008. 2

    work page 2008

  3. [3]

    Frozen in time: A joint video and image encoder for end-to-end retrieval

    Max Bain, Arsha Nagrani, G ¨ul Varol, and Andrew Zisser- man. Frozen in time: A joint video and image encoder for end-to-end retrieval. In ICCV, pages 1728–1738, 2021. 3

    work page 2021

  4. [4]

    Context recov- ery and knowledge retrieval: A novel two-stream framework for video anomaly detection

    Congqi Cao, Yue Lu, and Yanning Zhang. Context recov- ery and knowledge retrieval: A novel two-stream framework for video anomaly detection. IEEE Transactions on Image Processing, 33:1810–1825, 2024. 1

    work page 2024

  5. [5]

    Gramian multimodal representation learning and alignment

    Giordano Cicchetti, Eleonora Grassucci, Luigi Sigillo, and Danilo Comminiello. Gramian multimodal representation learning and alignment. In ICLR, 2025. 2, 3, 5, 6, 7

    work page 2025

  6. [6]

    Deepseek-r1: Incentivizing reasoning capa- bility in llms via reinforcement learning, 2025

    DeepSeek-AI. Deepseek-r1: Incentivizing reasoning capa- bility in llms via reinforcement learning, 2025. 4

    work page 2025

  7. [7]

    https://github.com/modelscope/diffsynth- studio, 2023

    DiffSynth-Studio. https://github.com/modelscope/diffsynth- studio, 2023. 4

    work page 2023

  8. [8]

    Oops! pre- dicting unintentional action in video

    Dave Epstein, Boyuan Chen, and Carl V ondrick. Oops! pre- dicting unintentional action in video. In CVPR, pages 919– 929, 2020. 2, 3, 6

    work page 2020

Show all 46 references
  1. [9]

    Mist: Multiple instance self-training framework for video anomaly detection

    Jia-Chang Feng, Fa-Ting Hong, and Wei-Shi Zheng. Mist: Multiple instance self-training framework for video anomaly detection. In CVPR, pages 14009–14018, 2021. 2

    2021

  2. [10]

    Cnvid-3.5 m: Build, fil- ter, and pre-train the large-scale public chinese video-text dataset

    Tian Gan, Qing Wang, Xingning Dong, Xiangyuan Ren, Liqiang Nie, and Qingpei Guo. Cnvid-3.5 m: Build, fil- ter, and pre-train the large-scale public chinese video-text dataset. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 14815– 1...

    2023

  3. [11]

    Tem- poral tessellation: A unified approach for video analysis

    Dotan Kaufman, Gil Levi, Tal Hassner, and Lior Wolf. Tem- poral tessellation: A unified approach for video analysis. In ICCV, pages 94–104, 2017. 3

    2017

  4. [12]

    Unifying visual-semantic embeddings with multimodal neu- ral language models

    Ryan Kiros, Ruslan Salakhutdinov, and Richard S Zemel. Unifying visual-semantic embeddings with multimodal neu- ral language models. arXiv preprint arXiv:1411.2539, 2014. 3

    2014 arXiv

  5. [13]

    Less is more: Clipbert for video-and-language learning via sparse sampling

    Jie Lei, Linjie Li, Luowei Zhou, Zhe Gan, Tamara L Berg, Mohit Bansal, and Jingjing Liu. Less is more: Clipbert for video-and-language learning via sparse sampling. In PCVPR, pages 7331–7341, 2021. 3

    2021

  6. [14]

    Selvaraju, Akhilesh Deepak Got- mare, Shafiq Joty, Caiming Xiong, and Steven Hoi

    Junnan Li, Ramprasaath R. Selvaraju, Akhilesh Deepak Got- mare, Shafiq Joty, Caiming Xiong, and Steven Hoi. Align be- fore fuse: Vision and language representation learning with momentum distillation. In NeurIPS, 2021. 3

    2021

  7. [15]

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In ICML, pages 12888–12900. PMLR, 2022. 3

    2022

  8. [16]

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

    Weixin Li, Vijay Mahadevan, and Nuno Vasconcelos. Anomaly detection and localization in crowded scenes.IEEE transactions on pattern analysis and machine intelligence , 36(1):18–32, 2013. 2

    2013

  9. [17]

    Fine-grained key-value mem- ory enhanced predictor for video representation learning

    Xiaojie Li, Jianlong Wu, Shaowei He, Shuo Kang, Yue Yu, Liqiang Nie, and Min Zhang. Fine-grained key-value mem- ory enhanced predictor for video representation learning. In ACM MM, pages 2264–2274, 2023. 3 7

    2023

  10. [18]

    Timestep embedding tells: It’s time to cache for video diffusion model

    Feng Liu, Shiwei Zhang, Xiaofeng Wang, Yujie Wei, Haonan Qiu, Yuzhong Zhao, Yingya Zhang, Qixiang Ye, and Fang Wan. Timestep embedding tells: It’s time to cache for video diffusion model. CVPR, 2025. 4

    2025

  11. [19]

    Ntu rgb+ d 120: A large- scale benchmark for 3d human activity understanding

    Jun Liu, Amir Shahroudy, Mauricio Perez, Gang Wang, Ling-Yu Duan, and Alex C Kot. Ntu rgb+ d 120: A large- scale benchmark for 3d human activity understanding. IEEE transactions on pattern analysis and machine intelligence , 42(10):2684–2701, 2019. 3

    2019

  12. [20]

    Fu- ture frame prediction for anomaly detection – a new baseline

    Wen Liu, Weixin Luo, Dongze Lian, and Shenghua Gao. Fu- ture frame prediction for anomaly detection – a new baseline. In CVPR, 2018. 2

    2018

  13. [21]

    Abnormal event detec- tion at 150 fps in matlab

    Cewu Lu, Jianping Shi, and Jiaya Jia. Abnormal event detec- tion at 150 fps in matlab. In ICCV, pages 2720–2727, 2013. 2

    2013

  14. [22]

    Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning

    Huaishao Luo, Lei Ji, Ming Zhong, Yang Chen, Wen Lei, Nan Duan, and Tianrui Li. Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning. Neu- rocomputing, 508:293–304, 2022. 2, 3, 5, 6

    2022

  15. [23]

    A revisit of sparse coding based anomaly detection in stacked rnn framework

    Weixin Luo, Wen Liu, and Shenghua Gao. A revisit of sparse coding based anomaly detection in stacked rnn framework. In ICCV, pages 341–349, 2017. 2

    2017

  16. [24]

    X-clip: End-to-end multi-grained con- trastive learning for video-text retrieval

    Yiwei Ma, Guohai Xu, Xiaoshuai Sun, Ming Yan, Ji Zhang, and Rongrong Ji. X-clip: End-to-end multi-grained con- trastive learning for video-text retrieval. In ACM MM, pages 638–647, 2022. 2, 3, 5, 6, 7

    2022

  17. [25]

    MULDE: Multiscale Log- Density Estimation via Denoising Score Matching for Video Anomaly Detection

    Jakub Micorek, Horst Possegger, Dominik Narnhofer, Horst Bischof, and Mateusz Kozi ´nski. MULDE: Multiscale Log- Density Estimation via Denoising Score Matching for Video Anomaly Detection. In CVPR, pages 18868–18877, 2024. 1, 2

    2024

  18. [26]

    End-to-end learning of visual representations from uncurated instruc- tional videos

    Antoine Miech, Jean-Baptiste Alayrac, Lucas Smaira, Ivan Laptev, Josef Sivic, and Andrew Zisserman. End-to-end learning of visual representations from uncurated instruc- tional videos. In CVPR, pages 9879–9889, 2020. 3

    2020

  19. [27]

    Learning memory-guided normality for anomaly detection

    Hyunjong Park, Jongyoun Noh, and Bumsub Ham. Learning memory-guided normality for anomaly detection. In CVPR,

  20. [28]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, pages 8748–8763. PmLR, 2021. 3, 6

    2021

  21. [29]

    Street scene: A new dataset and evaluation protocol for video anomaly detection

    Bharathkumar Ramachandra and Michael Jones. Street scene: A new dataset and evaluation protocol for video anomaly detection. In WACV, pages 2569–2578, 2020. 2

    2020

  22. [30]

    Deep-cascade: Cascading 3d deep neu- ral networks for fast anomaly detection and localization in crowded scenes

    Mohammad Sabokrou, Mohsen Fayyaz, Mahmood Fathy, and Reinhard Klette. Deep-cascade: Cascading 3d deep neu- ral networks for fast anomaly detection and localization in crowded scenes. IEEE Transactions on Image Processing , page 1992–2004, 2017. 2

    1992

  23. [31]

    Seed-thinking-v1

    ByteDance Seed, Yufeng Yuan, Yu Yue, Mingxuan Wang, Xiaochen Zuo, Jiaze Chen, Lin Yan, Wenyuan Xu, Chi Zhang, Xin Liu, et al. Seed-thinking-v1. 5: Advancing su- perb reasoning models with reinforcement learning. arXiv preprint arXiv:2504.13914, 2025. 4

    2025

  24. [32]

    Real-world anomaly detection in surveillance videos

    Waqas Sultani, Chen Chen, and Mubarak Shah. Real-world anomaly detection in surveillance videos. In CVPR, pages 6479–6488, 2018. 1, 2, 3, 5

    2018

  25. [33]

    Learning language-visual embedding for movie understanding with natural-language

    Atousa Torabi, Niket Tandon, and Leonid Sigal. Learning language-visual embedding for movie understanding with natural-language. arXiv preprint arXiv:1609.08124 , 2016. 3

    2016 arXiv

  26. [34]

    Wan: Open and advanced large-scale video generative models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianx- iao Yang, Jianyuan Zeng, Jiayu Wang, Jingfeng Zhang, Jin- gren Zhou, Jinkai Wang, Jixuan Chen, Kai Zhu, Kang Zhao, Keyu Yan, Lianghua Huang, Mengyang Feng, Ningyi Zhang, P...

    2025 arXiv

  27. [35]

    Align and tell: Boosting text-video re- trieval with local alignment and fine-grained supervision

    Xiaohan Wang, Linchao Zhu, Zhedong Zheng, Mingliang Xu, and Yi Yang. Align and tell: Boosting text-video re- trieval with local alignment and fine-grained supervision. IEEE Transactions on Multimedia, 25:6079–6089, 2022. 3

    2022

  28. [36]

    Rtq: Rethinking video-language under- standing based on image-text model

    Xiao Wang, Yaoyu Li, Tian Gan, Zheng Zhang, Jingjing Lv, and Liqiang Nie. Rtq: Rethinking video-language under- standing based on image-text model. In ACM MM, pages 557–566, 2023. 3

    2023

  29. [37]

    Weakly-supervised spatio-temporal anomaly detection in surveillance video

    Jie Wu, Wei Zhang, Guanbin Li, Wenhao Wu, Xiao Tan, Yingying Li, Errui Ding, and Liang Lin. Weakly-supervised spatio-temporal anomaly detection in surveillance video. In IJCAI, 2021. 2

    2021

  30. [38]

    An empirical study of frame selection for text-to-video retrieval

    Mengxia Wu, Min Cao, Yang Bai, Ziyin Zeng, Chen Chen, Liqiang Nie, and Min Zhang. An empirical study of frame selection for text-to-video retrieval. arXiv preprint arXiv:2311.00298, 2023. 3

    2023 arXiv

  31. [39]

    A deep one-class neural network for anomalous event detection in complex scenes

    Peng Wu, Jing Liu, and Fang Shen. A deep one-class neural network for anomalous event detection in complex scenes. IEEE Transactions on Neural Networks and Learning Sys- tems, page 1–14, 2019. 2

    2019

  32. [40]

    Not only look, but also listen: Learning multimodal violence detection under weak supervision

    Peng Wu, Jing Liu, Yujia Shi, Yujia Sun, Fangtao Shao, Zhaoyang Wu, and Zhiwei Yang. Not only look, but also listen: Learning multimodal violence detection under weak supervision. In ECCV, pages 322–339. Springer, 2020. 2

    2020

  33. [41]

    Toward video anomaly retrieval from video anomaly detection: New benchmarks and model.IEEE Transactions on Image Processing, 33:2213–2225, 2024

    Peng Wu, Jing Liu, Xiangteng He, Yuxin Peng, Peng Wang, and Yanning Zhang. Toward video anomaly retrieval from video anomaly detection: New benchmarks and model.IEEE Transactions on Image Processing, 33:2213–2225, 2024. 1, 2, 3, 5, 6

    2024

  34. [42]

    Qwen3 technical report

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chen- gen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, 8 Jianxin Yang,...

    2025 arXiv

  35. [43]

    A joint se- quence fusion model for video question answering and re- trieval

    Youngjae Yu, Jongseok Kim, and Gunhee Kim. A joint se- quence fusion model for video question answering and re- trieval. In ECCV, pages 471–487, 2018. 3

    2018

  36. [44]

    Towards surveillance video-and-language understanding: New dataset baselines and challenges

    Tongtong Yuan, Xuange Zhang, Kun Liu, Bo Liu, Chen Chen, Jian Jin, and Zhenzhen Jiao. Towards surveillance video-and-language understanding: New dataset baselines and challenges. In CVPR, pages 22052–22061, 2024. 1, 2, 3, 5

    2024

  37. [45]

    Generative cooper- ative learning for unsupervised video anomaly detection

    M Zaigham Zaheer, Arif Mahmood, M Haris Khan, Mat- tia Segu, Fisher Yu, and Seung-Ik Lee. Generative cooper- ative learning for unsupervised video anomaly detection. In CVPR, pages 14744–14754, 2022. 2

    2022

  38. [46]

    Multi-grained vi- sion language pre-training: Aligning texts with visual con- cepts

    Yan Zeng, Xinsong Zhang, and Hang Li. Multi-grained vi- sion language pre-training: Aligning texts with visual con- cepts. ICML, 2022. 3 9

    2022

Pith tools

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