Pith. sign in

REVIEW 4 major objections 6 minor 58 references

Zero-Shot Crowd Behavior Recognition

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read Crowd behaviors with zero training videos become recognizable

desk verdict First multi-label ZSL for crowd behavior with a useful violence transfer, but the conditional probability is mis-normalized and the learned co-occurrence map never sees negative pairs. read the letter →

arxiv 1908.05877 v1 pith:AQD4VQZJ submitted 2019-08-16 cs.CV

classification cs.CV
keywords zero-shotlearningcrowdbehaviorrecognitionmulti-labelclassificationattributeco-occurrenceword-vectorembeddingsvideosurveillanceviolencedetectionbilinearmapping
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 establish that a video system can recognize crowd behaviors it has never been trained on, such as violence, by combining confidence scores for known behaviors with predicted relations between known and novel behaviors. Since crowd videos are inherently multi-label, the paper argues zero-shot recognition should exploit co-occurrence context rather than treat each behavior independently. Its proposal is a bilinear map $M$, learned from word-vector embeddings and annotated visual co-occurrence of known attribute pairs, that predicts co-occurrence for pairs involving an unseen attribute. If correct, novel-behavior detection becomes possible without new annotations: the paper reports a zero-shot AUC of 0.69 on the WWW dataset versus 0.65 for the best prior zero-shot baseline, and a transferred violence detector reaching 87.22 AUC, above a supervised baseline's 85.00.

What carries the argument

The load-bearing object is the bilinear map $M$ of Eq. (6), a matrix that takes a pair of word vectors $(v_i, v_j)$ and outputs the expected log visual co-occurrence $\log c_{ij}$ of the two attributes. It is fit by weighted ridge regression against the co-occurrence matrix $C = Y_S Y_S^\top$ computed from multi-label annotations of known attributes, with a weighting function that down-weights rare pairs. The same $M$ is then used in Eq. (8) to score conditional probabilities $p(y^*_q \mid y_p)$ for pairs that include a novel attribute, and those conditionals are marginalized over known-attribute confidences in Eq. (1) to predict novel labels. The map is the mechanism that lets word-vector semantics be corrected by and aligned with visual context, and then extrapolated to unseen attribute pairs.

What would settle it

Annotate a fresh set of crowd videos for the held-out novel attributes, compute their empirical co-occurrence with known attributes, and compare these counts against the conditional probabilities predicted by $M$; if the rank correlation is near zero or negative, the extrapolation at the core of the paper collapses.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that multi-label zero-shot crowd behavior recognition can be solved by marginalizing known-attribute predictions through estimated known-to-novel attribute relations. Its model, CoCAZSL, learns a probabilistic classifier for known attributes, then forms conditional probabilities between novel and known attributes from a bilinear mapping $M$ that predicts visual co-occurrence from word vectors. Setting $M$ to the identity recovers text-only relations; learning $M$ from annotated co-occurrence corrects cases where linguistic similarity mismatches visual context, such as indoor and outdoor. Experimentally, the paper reports that this context-aware marginalization improves label-based AUC and AP over four existing zero-shot models on the WWW dataset, and that transferring the learned relations to violence detection yields AUC 87.22, above the supervised ViF baseline's 85.00. The claim is that contextual co-occurrence, transferred through a learned word-pair map, is what makes zero-shot crowd behavior recognition work.

Load-bearing premise

The single bilinear rule learned from pairs of known attributes keeps predicting co-occurrence accurately for pairs that include a novel attribute.

Editorial extensions

If this is right

  • Zero-shot crowd attribute prediction is feasible: on the WWW dataset all tested models beat random guessing, and the context-aware model reports the highest label-based AUC and AP.
  • Adding visual co-occurrence context to a text-only semantic relation improves prediction: CoCAZSL reaches 0.69 AUC versus 0.65 for TexCAZSL.
  • The learned relations transfer across domains: without any violence training examples, CoCAZSL reaches 87.22 AUC on Violent Flow, above the supervised ViF baseline's 85.00.
  • Multi-label context is what supplies the missing statistics: because novel attributes cannot have annotated co-occurrence, the bilinear map generates those statistics from word pairs and known-attribute annotations.

Reading between the lines

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

  • A natural extension the paper leaves implicit is learning separate co-occurrence maps for the 'where', 'who', and 'why' attribute groups, which could test whether group-specific context extrapolates better than a single global map.
  • The same word-pair to co-occurrence transfer could apply to other multi-label zero-shot problems, such as tagging images with novel objects or actions, whenever word vectors and partial label co-occurrence are available.
  • The indoor/outdoor failure of text-only relations suggests a general caution: any text-based semantic prior should be calibrated against visual annotation statistics before being trusted for visual prediction.
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

4 major / 6 minor

Summary. This manuscript proposes CoCAZSL, a zero-shot multi-label crowd behavior recognition method. The model first trains probabilistic classifiers for known attributes, then estimates relations from known to novel attributes via a bilinear map M that predicts visual co-occurrence from pairs of word vectors, and finally marginalizes known-attribute predictions through these relations in an IAP-style formulation given by Eq. (1). The authors also present a text-only variant (TexCAZSL). Experiments on the WWW crowd video dataset report improved label-based AUC and AP over four ZSL baselines, and a transfer experiment to Violence Flow suggests competitive zero-shot violence detection. The central claim is that modeling multi-attribute co-occurrence context improves zero-shot crowd behavior recognition and generalizes to novel behaviors cross-domain.

Significance. If the results hold, the paper makes a useful contribution by introducing zero-shot learning to multi-label crowd behavior recognition and by showing that co-occurrence context learned from known attributes can be transferred to novel attributes. The idea of learning a pairwise mapping from text embeddings to visual co-occurrence is interesting and the WWW dataset provides a challenging testbed. The reported AUC gains (0.69 vs 0.65 for TexCAZSL) and the cross-domain violence transfer are compelling. However, the two technical issues detailed below concern the probabilistic validity of the conditioning step and the mechanism by which visual compatibility is learned; both are load-bearing for the stated claims.

major comments (4)
  1. [§3.2.2, Eq. (8)] The quantity defined in Eq. (8) is not the conditional probability p(y*_q | y_p) used in Eq. (1). The denominator sums over the conditioning variable p (known attributes), producing a distribution over p for each fixed q, whereas a conditional p(y*_q | y_p) should sum over q to 1. This is also inconsistent with Eq. (5), where the normalization is over the conditioned variable j. Consequently, the marginalization in Eq. (1) does not yield calibrated probabilities over novel labels, and the relative scale of scores for different q is arbitrary. This scale arbitrariness affects the example-based metrics (e.g., example-based AP and ranking loss) though not the per-label AUC. Please clarify whether Eq. (8) is intended as an unnormalized affinity; if so, the probabilistic graphical model in §3.1 and the interpretation of Eq. (1) as a marginal conditional probability need to be revised.
  2. [§3.2.2, Eqs. (6)-(7)] The regression in Eq. (6) is trained only on known attribute pairs with positive co-occurrence because the weight function in Eq. (7) gives w(0)=0 for alpha>0. The stated motivation for using visual co-occurrence is that textually similar pairs such as "indoor" and "outdoor" have zero visual co-occurrence and should be suppressed, but the objective never presents such incompatible pairs to M. Thus, the learned mapping has no pressure to output low values for non-co-occurring pairs, and the improvement of CoCAZSL over TexCAZSL in Table 2 (AUC 0.69 vs 0.65) cannot be attributed to the "visual compatibility" mechanism as described. Please either include zero-co-occurrence pairs with a non-zero penalty in Eq. (6), or provide direct evidence (e.g., predicted co-occurrence for held-out known-known pairs) that M extrapolates to incompatible unseen pairs.
  3. [Table 2, §4.1.2] The reported gains of CoCAZSL over TexCAZSL (AUC 0.69 vs 0.65; label-based AP 0.27 vs 0.24) are the central empirical evidence for the contribution, but the paper gives no variance or significance information across the 50 random splits. Since the difference is modest and the baselines are also within 0.02-0.03 of each other, please report per-split mean±std or paired tests across the 50 splits to show that the improvement is systematic rather than due to a few favourable splits.
  4. [§4.2, Table 3] The claim that the zero-shot model "beats the fully supervised Linear SVM with ViF feature in AUC (87.22 vs 85.00)" compares CoCAZSL using ITF with a supervised baseline using ViF. The same-feature supervised ITF baseline achieves 98.72 AUC, which is substantially higher. Please reframe the claim to acknowledge the feature mismatch and, if possible, add a supervised ViF baseline or a zero-shot model using the same features to make the comparison apples-to-apples.
minor comments (6)
  1. [§3.1, Eq. (1)] The second line "p( ˜y ˜p|x)" appears to be a typo; it should presumably read p(\tilde{y}_{\tilde{p}}|x).
  2. [§3.2.2, Eq. (9)] The gradient in Eq. (9) uses f(c_ij) but the loss in Eq. (6) uses w(c_ij); please make the notation consistent.
  3. [§4.1.1] The parameter setting "λ = 1−3" should likely be "λ = 10^{-3}"; please clarify.
  4. [Table 3] The supervised baseline is cited as "[18]" in the table, but the text refers to "the best performance of [19]"; the citation appears inconsistent.
  5. [Fig. 9] The caption contains a typo: "Importantce" should be "Importance".
  6. [Section 2.2] In the phrase "Indirect Attribute Predici ton (IAP)", "Predici ton" should be "Prediction".

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the bilinear co-occurrence map is fitted on known attributes only and is genuinely extrapolated to novel attributes and external benchmarks.

full rationale

The paper's derivation chain is self-contained and not circular. The key predictive quantity is p(y*_q | x) = sum_p p(y*_q | y_p) p(y_p | x) in Eq. (1). The known-to-novel conditional p(y*_q | y_p) is obtained in Eq. (8) by exponentiating v_q^T M v_p, where the bilinear map M is learned in Eq. (6) from word-vector pairs and visual co-occurrence statistics of the 85 known attributes only. Novel attribute annotations and test labels never enter the estimation of M; the only bridge to novel attributes is the external Google News word-vector embedding. Thus the reported WWW and Violence Flow results are not forced by construction: Eq. (6) solves a genuine regression on seen attribute pairs, and Eq. (8) applies the fitted map to unseen attribute pairs, which is extrapolation rather than a refitting of the target quantity. The weight function in Eq. (7) does zero out zero-co-occurrence pairs, so the model is not trained on negative visual co-occurrence evidence; this is a limitation of the extrapolation assumption, not a circular reduction, because the evaluated novel-attribute predictions are still not derived from the labels being predicted. Self-citations such as [12], [41], [52], and [53] are used as compared baselines, dataset references, or related prior work; none is invoked as an unverified uniqueness theorem or as the sole justification of the central transfer step. The central claim therefore has independent empirical content and no circular step was found.

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

The method rests on two main assumptions: novel attributes interact with video content only through known-attribute classifier outputs (Eq. 1), and visual co-occurrence of attribute pairs is a global bilinear function of word vectors that generalizes to unseen pairs (Eq. 6, used in Eq. 8). The co-occurrence weighting function introduces several unspecified hyperparameters. No new entities are postulated; the 'violence' attribute is an existing label in the target dataset.

free parameters (4)
  • lambda (ridge regularization for bilinear map M) = printed as '1-3', likely 1e-3
    Regularizer in Eq. (6) chosen by hand to avoid overfitting; value not precisely stated.
  • gamma (temperature in text softmax) = 0.1
    Cross-validated on the WWW training split for Eq. (4); controls sharpness of text-based conditional weights.
  • alpha and Cmax (co-occurrence weighting function) = not reported
    Parameters of w(c_ij) in Eq. (7), inherited from Mensink et al. [28]; no values given, affecting the regression loss weighting.
  • SVM slack C for known-attribute classifiers = 1
    Fixed slack parameter for the 85 per-attribute linear SVMs; chosen without tuning.
assumptions (4)
  • domain assumption Novel attribute scores are obtained by marginalizing over known attributes: p(y*_q | x) = sum_p p(y*_q | y_p) p(y_p | x).
    Eq. (1). Implicitly assumes the video's visual content affects novel attributes only through known-attribute classifier confidences; a novel attribute with no detectable known-attribute evidence cannot be recognized.
  • ad hoc to paper Visual co-occurrence of attribute pairs is a bilinear function of word vectors via a single global matrix M: exp(v_i^T M v_j) approximates c_ij.
    Eq. (6). The paper postulates this mapping and relies on it to extrapolate to pairs involving novel attributes that have no co-occurrence statistics. No kernel or nonlinear form is considered.
  • domain assumption Pre-trained skip-gram word vectors (Google News, 300-d) capture semantic relations that transfer to crowd-attribute co-occurrence.
    Section 3.2.1. The attribute names are mapped directly into this unsupervised text space; no domain-specific word embeddings are learned.
  • ad hoc to paper The normalized quantity in Eq. (8) can be used as p(y*_q | y_p) despite normalizing over the conditioning variable.
    Section 3.2.2. As written, sum_p p(y*_q | y_p) = 1 for each q, so the expression is not a conditional distribution over novel attributes; the paper nonetheless inserts it into the Bayes-style marginal in Eq. (1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zero-Shot Crowd Behavior Recognition." pith.science (2026). https://pith.science/paper/AQD4VQZJ

@misc{pith2026190805877,
  author       = {Pith},
  title        = {Pith review of: Zero-Shot Crowd Behavior Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AQD4VQZJ}},
  note         = {Machine review of arXiv:1908.05877}
}
read the original abstract

Understanding crowd behavior in video is challenging for computer vision. There have been increasing attempts on modeling crowded scenes by introducing ever larger property ontologies (attributes) and annotating ever larger training datasets. However, in contrast to still images, manually annotating video attributes needs to consider spatiotemporal evolution which is inherently much harder and more costly. Critically, the most interesting crowd behaviors captured in surveillance videos (e.g., street fighting, flash mobs) are either rare, thus have few examples for model training, or unseen previously. Existing crowd analysis techniques are not readily scalable to recognize novel (unseen) crowd behaviors. To address this problem, we investigate and develop methods for recognizing visual crowd behavioral attributes without any training samples, i.e., zero-shot learning crowd behavior recognition. To that end, we relax the common assumption that each individual crowd video instance is only associated with a single crowd attribute. Instead, our model learns to jointly recognize multiple crowd behavioral attributes in each video instance by exploring multiattribute cooccurrence as contextual knowledge for optimizing individual crowd attribute recognition. Joint multilabel attribute prediction in zero-shot learning is inherently nontrivial because cooccurrence statistics does not exist for unseen attributes. To solve this problem, we learn to predict cross-attribute cooccurrence from both online text corpus and multilabel annotation of videos with known attributes. Our experiments show that this approach to modeling multiattribute context not only improves zero-shot crowd behavior recognition on the WWW crowd video dataset, but also generalizes to novel behavior (violence) detection cross-domain in the Violence Flow video dataset.

Figures

Figures reproduced from arXiv: 1908.05877 by the authors.

Figure 1
Figure 1. A thumbnail visualisation and a summary on the popularities of all 94 at [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. In model training, we learn word-vector representations of training attributes [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. A probablistic graphical representation of a context-aware multi-label zero [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Examples of all attributes in the WWW crowd video dataset [41] [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Statistics of the dataset split for our experiments on the [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Illustration of crowd videos ranked in accordance with prediction scores [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Examples of zero-shot multi-label attribute prediction. Bars under each im [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: Example frames of violence flow dataset [19]. [PITH_FULL_IMAGE:figures/full_fig_p022_8.png]
Figure 9
Figure 9. Figure 9: Importantce of known attributes w.r.t. novel event/attributes. The fontsize of [PITH_FULL_IMAGE:figures/full_fig_p026_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 55 canonical work pages

  1. [1]

    Evaluation of output embeddings for fine-grained image classification

    Zeynep Akata, Scott Reed, Daniel Walter, Honglak Lee, and Bernt Schiele. Evaluation of output embeddings for fine-grained image classification. In CVPR, 2015

  2. [2]

    Exploring synonyms as context in zero-shot action recognition

    Ioannis Alexiou, Tao Xiang, and Shaogang Gong. Exploring synonyms as context in zero-shot action recognition. In ICIP, 2016

  3. [3]

    Floor fields for tracking in high density crowd scenes

    Saad Ali and Mubarak Shah. Floor fields for tracking in high density crowd scenes. In ECCV, 2008

  4. [4]

    Modelling crowd scenes for event detection

    Ernesto L Andrade, Scott Blunsden, and Robert B Fisher. Modelling crowd scenes for event detection. In ICPR, 2006

  5. [5]

    Learning multi-label scene classification

    Matthew R Boutell, Jiebo Luo, Xipeng Shen, and Christopher M Brown. Learning multi-label scene classification. Pattern recognition, 2004

  6. [6]

    LIBSVM: A library for support vector machines

    Chih-Chung Chang and Chih-Jen Lin. LIBSVM: A library for support vector machines. ACM Transactions on Intelligent Systems and Technology, 2011. Software available at http:// www.csie.ntu.edu.tw/˜cjlin/libsvm

  7. [7]

    Feature mining for localised crowd counting

    Ke Chen, Chen Change Loy, Shaogang Gong, and Tao Xiang. Feature mining for localised crowd counting. In BMVC, 2012

  8. [8]

    Histograms of oriented gradients for human detection

    Navneet Dalal and Bill Triggs. Histograms of oriented gradients for human detection. In CVPR, 2005

Show all 58 references
  1. [9]

    Christiane Fellbaum. WordNet. Wiley Online Library

  2. [10]

    Devise: A deep visual-semantic embedding model

    Andrea Frome, Greg S Corrado, Jon Shlens, Samy Bengio, Jeff Dean, Tomas Mikolov, et al. Devise: A deep visual-semantic embedding model. In NIPS, 2013

  3. [11]

    Transductive Multi- view Zero-Shot Learning

    Yanwei Fu, Timothy M Hospedales, Tao Xiang, and Shaogang Gong. Transductive Multi- view Zero-Shot Learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015

  4. [12]

    Trans- ductive Multi-Label Zero-shot Learning

    Yanwei Fu, Yongxin Yang, Timothy M Hospedales, Tao Xiang, and Shaogang Gong. Trans- ductive Multi-Label Zero-shot Learning. In BMVC, 2014

  5. [13]

    Multilabel classification via calibrated label ranking

    Johannes F ¨urnkranz, Eyke H ¨ullermeier, Eneldo Loza Menc´ıa, and Klaus Brinker. Multilabel classification via calibrated label ranking. Machine learning, 2008

  6. [14]

    Exploring semantic inter-class relationships (sir) for zero-shot action recognition

    Chuang Gan, Ming Lin, Yi Yang, Yueting Zhuang, and Alexander G.Hauptmann. Exploring semantic inter-class relationships (sir) for zero-shot action recognition. In AAAI, 2015

  7. [15]

    Devnet: A deep event network for multimedia event detection and evidence recounting

    Chuang Gan, Naiyan Wang, Yi Yang, Dit-Yan Yeung, and Alex G Hauptmann. Devnet: A deep event network for multimedia event detection and evidence recounting. In CVPR, 2015

  8. [16]

    Collective multi-label classification

    Nadia Ghamrawi and Andrew McCallum. Collective multi-label classification. In ACM CIKM, 2005

  9. [17]

    Security and surveillance

    Shaogang Gong, Chen Change Loy, and Tao Xiang. Security and surveillance. In Moeslund, Hilton, Kruger, and Sigal, editors,Visual Analysis of Humans, pages 455–472. Springer, 2011

  10. [18]

    Efficient max-margin multi-label classification with applications to zero-shot learning

    Bharath Hariharan, SVN Vishwanathan, and Manik Varma. Efficient max-margin multi-label classification with applications to zero-shot learning. Machine learning, 2012

  11. [19]

    Violent flows: Real-time detection of violent crowd behavior

    Tal Hassner, Yossi Itcher, and Orit Kliper-Gross. Violent flows: Real-time detection of violent crowd behavior. In CVPR Workshop, 2012

  12. [20]

    Yu-Gang Jiang, Guangnan Ye, Shih-Fu Chang, Daniel P. W. Ellis, and Alexander C. Loui. Consumer video understanding: a benchmark database and an evaluation of human and ma- chine performance. In ICMR, 2011. 28 Xun Xu, Shaogang Gong and Timothy M. Hospedales

  13. [21]

    Anomaly detection in extremely crowded scenes using spatio- temporal motion pattern models

    Louis Kratz and Ko Nishino. Anomaly detection in extremely crowded scenes using spatio- temporal motion pattern models. In CVPR, 2009

  14. [22]

    Hmdb: A large video database for human motion recognition

    H Kuehne, H Jhuang, E Garrote, T Poggio, and T Serre. Hmdb: A large video database for human motion recognition. In ICCV, 2011

  15. [23]

    Lampert, H

    C.H. Lampert, H. Nickisch, and S. Harmeling. Learning to detect unseen object classes by between-class attribute transfer. In CVPR, 2009

  16. [24]

    Beyond bags of features: Spatial pyra- mid matching for recognizing natural scene categories

    Svetlana Lazebnik, Cordelia Schmid, and Jean Ponce. Beyond bags of features: Spatial pyra- mid matching for recognizing natural scene categories. CVPR, 2006

  17. [25]

    Learning behavioural context

    Jian Li, Shaogang Gong, and Tao Xiang. Learning behavioural context. International journal of computer vision, 2012

  18. [26]

    Multi-label image classification with a probabilistic label enhancement model

    Xin Li, Feipeng Zhao, and Yuhong Guo. Multi-label image classification with a probabilistic label enhancement model. In UAI, 2014

  19. [27]

    Crowd counting and profil- ing: Methodology and evaluation

    Chen Change Loy, Ke Chen, Shaogang Gong, and Tao Xiang. Crowd counting and profil- ing: Methodology and evaluation. In Ali, Nishino, Manocha, and Shah, editors, Modeling, Simulation and Visual Analysis of Crowds. Springer, December 2013

  20. [28]

    COSTA: Co-occurrence statistics for zero-shot classification

    Thomas Mensink, Efstratios Gavves, and Cees G M Snoek. COSTA: Co-occurrence statistics for zero-shot classification. In CVPR, 2014

  21. [29]

    Distributed Rep- resentations of Words and Phrases and their Compositionality

    Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. Distributed Rep- resentations of Words and Phrases and their Compositionality. In NIPS, 2013

  22. [30]

    Modeling temporal structure of decom- posable motion segments for activity classification

    Juan Carlos Niebles, Chih Wei Chen, and Li Fei-Fei. Modeling temporal structure of decom- posable motion segments for activity classification. In ECCV, 2010

  23. [31]

    Modeling the shape of the scene: A holistic representation of the spatial envelope

    Aude Oliva and Antonio Torralba. Modeling the shape of the scene: A holistic representation of the spatial envelope. International Journal of Computer Vision, 2001

  24. [32]

    Deepid-net: multi-stage and deformable deep convolutional neural networks for object detection

    Wanli Ouyang, Ping Luo, Xingyu Zeng, Shi Qiu, Yonglong Tian, Hongsheng Li, Shuo Yang, Zhe Wang, Yuanjun Xiong, Chen Qian, et al. Deepid-net: multi-stage and deformable deep convolutional neural networks for object detection. arXiv preprint arXiv:1409.3505, 2014

  25. [33]

    Glove: Global vectors for word representation

    Jeffrey Pennington, Richard Socher, and Christopher D Manning. Glove: Global vectors for word representation. In EMNLP, 2014

  26. [34]

    Improving the Fisher kernel for large-scale image classification

    Florent Perronnin, Jorge S ´anchez, and Thomas Mensink. Improving the Fisher kernel for large-scale image classification. In ECCV, 2010

  27. [35]

    Cor- relative multi-label video annotation

    Guo-Jun Qi, Xian-Sheng Hua, Yong Rui, Jinhui Tang, Tao Mei, and Hong-Jiang Zhang. Cor- relative multi-label video annotation. In ACM Multimedia, 2007

  28. [36]

    Data-driven crowd anal- ysis in videos

    Mikel Rodriguez, Josef Sivic, Ivan Laptev, and Jean-Yves Audibert. Data-driven crowd anal- ysis in videos. In ICCV, 2011

  29. [37]

    Evaluating knowledge transfer and zero- shot learning in a large-scale setting

    Marcus Rohrbach, Michael Stark, and Bernt Schiele. Evaluating knowledge transfer and zero- shot learning in a large-scale setting. In CVPR, 2011

  30. [38]

    What helps where - and why? Semantic relatedness for knowledge transfer

    Marcus Rohrbach, Michael Stark, Gy ¨orgy Szarvas, Iryna Gurevych, and Bernt Schiele. What helps where - and why? Semantic relatedness for knowledge transfer. In CVPR, 2010

  31. [39]

    An embarrassingly simple approach to zero-shot learning

    Bernardino Romera-Paredes. An embarrassingly simple approach to zero-shot learning. In ICML, 2015

  32. [40]

    Scene understanding by statistical mod- eling of motion patterns

    Imran Saleemi, Lance Hartung, and Mubarak Shah. Scene understanding by statistical mod- eling of motion patterns. In CVPR, 2010

  33. [41]

    Deeply learned attributes for crowded scene understanding

    Jing Shao, Kai Kang, Chen Change Loy, and Xiaogang Wang. Deeply learned attributes for crowded scene understanding. In CVPR, 2015

  34. [42]

    Scene-independent group profiling in crowd

    Jing Shao, Chen Loy, and Xiaogang Wang. Scene-independent group profiling in crowd. In CVPR, 2014

  35. [43]

    Zero-shot learning through cross-modal transfer

    Richard Socher and Milind Ganjoo. Zero-shot learning through cross-modal transfer. In NIPS, 2013

  36. [44]

    Ucf101: A dataset of 101 human actions classes from videos in the wild

    Khurram Soomro, A R Zamir, and Mubarak Shah. Ucf101: A dataset of 101 human actions classes from videos in the wild. arXiv preprint arXiv:1212.0402, 2012

  37. [45]

    Random k-labelsets: An ensemble method for multilabel classification

    Grigorios Tsoumakas and Ioannis Vlahavas. Random k-labelsets: An ensemble method for multilabel classification. In ECML, 2007

  38. [46]

    Parametric mixture models for multi-labeled text

    Naonori Ueda and Kazumi Saito. Parametric mixture models for multi-labeled text. In NIPS, 2002. Zero-Shot Crowd Behaviour Recognition 29

  39. [47]

    A sequential topic model for mining recurrent activities from long term video logs

    Jagannadan Varadarajan, Remi Emonet, and Jean-Marc Odobez. A sequential topic model for mining recurrent activities from long term video logs. International Journal of Computer Vision, 2013

  40. [48]

    Action recognition by dense trajectories

    Heng Wang, Alexander Kl ¨aser, Cordelia Schmid, and Cheng-Lin Liu. Action recognition by dense trajectories. In CVPR, 2011

  41. [49]

    A robust and efficient video representation for action recog- nition

    Heng Wang, Dan Oneata, Jakob Verbeek, Cordelia Schmid, Heng Wang, Dan Oneata, Jakob Verbeek, and Cordelia Schmid A. A robust and efficient video representation for action recog- nition. International Journal of Computer Vision, 2015

  42. [50]

    Unsupervised activity perception in crowded and complicated scenes using hierarchical bayesian models

    Xiaogang Wang, Xiaoxu Ma, and W Eric L Grimson. Unsupervised activity perception in crowded and complicated scenes using hierarchical bayesian models. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009

  43. [51]

    X. Xu, T. M. Hospedales, and S. Gong. Discovery of shared semantic spaces for multi-scene video query and summarization. IEEE Transactions on Circuits and Systems for Video Tech- nology, 2016

  44. [52]

    Cross-domain traffic scene understanding by motion model transfer

    Xun Xu, Shaogang Gong, and Timothy Hospedales. Cross-domain traffic scene understanding by motion model transfer. In Proceedings of the 4th ACM/IEEE International Workshop on ARTEMIS, 2013

  45. [53]

    Semantic embedding space for zero-shot action recognition

    Xun Xu, Timothy Hospedales, and Shaogang Gong. Semantic embedding space for zero-shot action recognition. In ICIP, 2015

  46. [54]

    Ml-knn: A lazy learning approach to multi-label learning

    Min-Ling Zhang and Zhi-Hua Zhou. Ml-knn: A lazy learning approach to multi-label learning. Pattern recognition, 2007

  47. [55]

    A review on multi-label learning algorithms

    Min-Ling Zhang and Zhi-Hua Zhou. A review on multi-label learning algorithms. IEEE Transactions on Knowledge and Data Engineering, 2014

  48. [56]

    Rotation-invariant image and video description with local binary pattern features

    Guoying Zhao, Timo Ahonen, Ji ˇr´ı Matas, and Matti Pietik¨ainen. Rotation-invariant image and video description with local binary pattern features. IEEE Transactions on Image Processing, 2012

  49. [57]

    Tracking using motion patterns for very crowded scenes

    Xuemei Zhao, Dian Gong, and G ´erard Medioni. Tracking using motion patterns for very crowded scenes. In ECCV, 2012

  50. [58]

    Coherent filtering: detecting coherent motions from crowd clutters

    Bolei Zhou, Xiaoou Tang, and Xiaogang Wang. Coherent filtering: detecting coherent motions from crowd clutters. In ECCV, 2012

Pith tools

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