Pith. sign in

REVIEW 4 major objections 6 minor 41 references

Shopformer: Transformer-Based Framework for Detecting Shoplifting via Human Pose

T0 review · 4 major / 6 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 Shopformer is a transformer framework that detects shoplifting from human pose sequences instead of raw video, and on the PoseLift dataset it reaches an AUC-ROC of 69.15%, ahead of three pose-based anomaly detectors.

desk verdict Shopformer's claimed SOTA is an artifact of metric choice and test-set tuning; the architecture is coherent but the evidence doesn't support the conclusion. read the letter →

arxiv 2504.19970 v1 pith:V7S7LLRM submitted 2025-04-28 cs.CV

Narges Rashvand , Ghazal Alinezhad Noghre , Armin Danesh Pazho , Babak Rahimi Ardabili , Hamed Tabkhi This is my paper · ORCID
classification cs.CV
keywords shopliftingdetectionhumanposeanalysistransformergraphconvolutionalautoencoderanomalyprivacy-preservingsurveillancesequencetokenizationreconstructionerror
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 argues that shoplifting can be recognized from the motion of a person's joints alone, without looking at what they are holding or what the store looks like. It introduces Shopformer, a two-stage model: a graph-convolutional autoencoder compresses pose sequences into tokens, and a transformer encoder-decoder learns to reconstruct normal shopping motion. During testing, reconstruction error becomes the shoplifting score, with worse reconstructions indicating more abnormal movement. On the PoseLift dataset, the best configuration reaches 69.15% AUC-ROC, outperforming STG-NF, TSGAD, and GEPC. If the result holds, surveillance systems could flag suspected theft using privacy-preserving skeletal data rather than identifiable video frames.

What carries the argument

The load-bearing object is the GCAE-encoder tokenizer: a graph-convolutional autoencoder whose encoder is trained to reconstruct pose sequences and is then frozen and repurposed to emit compact spatio-temporal tokens. Each token has dimensionality $C \times K$, where $C$ is the number of output channels and $K$ the number of keypoints; the paper ablated token counts from 1 to 12 and embedding sizes from 72 to 1152, finding two tokens of size 144 to be best. The transformer module is a standard encoder-decoder with self-attention, positional encoding, and a reconstruction (MSE) objective. Because it is trained only on normal pose sequences, its reconstruction error on new sequences is interpreted as a normality score that separates shoplifting from normal browsing.

What would settle it

On the PoseLift test split, compute the reconstruction MSE of the trained Shopformer for each of the 1,500 shoplifting frames and 2,221 normal frames, plot the two histograms, and measure the separation; if the shoplifting scores are not consistently higher than the normal scores, for example if the best threshold yields an AUC-ROC near 0.5, the reconstruction-error premise is refuted. A sharper version is to repeat training with several random seeds and report the spread of AUC-ROC values; if the 69.15% figure is not stable, the claimed gain is not yet established.

Watch

Extended reading notes

Core claim

The paper's central claim is that pose-sequence-level transformer modeling, paired with a graph-convolutional autoencoder (GCAE) tokenizer, is an effective and previously unexplored way to detect shoplifting. Shopformer is trained in two stages. First, a spatio-temporal graph autoencoder is trained to reconstruct pose sequences, and its encoder is frozen. Second, that frozen encoder turns each pose window into a small number of tokens, which a transformer encoder-decoder is trained to reconstruct, using only normal shopping sequences so that it learns typical motion. At inference, the mean squared error between input tokens and reconstructed tokens serves as the anomaly score: normal motion reconstructs well, while shoplifting motion does not. The paper reports that the best Shopformer configuration, with two tokens per window and a 144-dimensional token embedding, achieves 69.15% AUC-ROC on the PoseLift test set, compared with 67.46% for STG-NF, 63.35% for TSGAD, and 60.61% for GEPC. It therefore claims that its tokenization strategy and transformer architecture are what drive the gain.

Load-bearing premise

The model separates shoplifting from normal behavior only if normal pose sequences reconstruct with low mean squared error and shoplifting pose sequences with high error; if the two score distributions overlap even at the best threshold, the whole detection mechanism has no working signal.

Editorial extensions

If this is right

  • If Shopformer's claim is correct, shoplifting detection can operate on skeletal keypoints alone, eliminating the need to store or analyze identifiable video frames in retail surveillance.
  • The two-stage design means the pose-encoding tokenizer can be trained on unlabeled normal footage and reused for other behavior-based anomaly tasks.
  • A two-token-per-window configuration shows that aggressive temporal compression is not only cheaper but, in this setting, more discriminative than richer tokenizations.
  • The final chosen configuration has two transformer layers, two attention heads, and a feed-forward dimensionality of 64, which the paper identifies as a lightweight real-time trade-off.

Reading between the lines

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

  • A testable extension is to apply the same frozen pose tokenizer to other retail behaviors, such as queue jumping or fighting, to see whether the reconstruction-error score separates each behavior from normal motion rather than only shoplifting.
  • The ablation pattern, best at two tokens and degrading at three or four, suggests the tokenizer already packs the relevant motion into a very small latent; one could probe whether those latent dimensions correspond to semantically interpretable motion components, such as hand-to-pocket transitions.
  • Cross-dataset generalization is unexamined here; a next step would be to train on PoseLift and test on pose sequences from a different store layout or camera angle to measure whether the model learns shoplifting motion itself or dataset-specific quirks.
  • Because the normality score is a single per-window MSE, one could fuse frame-level scores with a temporal smoothing filter to produce incident-level alarms; the 43 annotated shoplifting incidents in the test set are few enough that such a fusion may materially change precision.
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 paper introduces Shopformer, a two-stage transformer-based framework for shoplifting detection from 2D human pose sequences. A graph convolutional autoencoder is trained on normal pose sequences; its frozen encoder is then used as a tokenizer that converts pose windows into compact embeddings. A transformer encoder-decoder is trained to reconstruct these tokens from normal sequences only, and the reconstruction MSE is used as the anomaly score at test time. The model is evaluated on the PoseLift dataset (53,353 normal training frames; test set with 1,500 shoplifting frames and 2,221 normal frames) and compared with STG-NF, TSGAD, and GEPC using AUC-ROC, AUC-PR, and EER. The authors report the highest AUC-ROC (69.15%) among the compared methods and claim that this confirms the effectiveness of their tokenization strategy and transformer architecture.

Significance. Shoplifting detection from pose data is a practical and privacy-preserving direction, and the paper makes a useful contribution by proposing a GCAE-based tokenization scheme for a transformer architecture. The public release of code is a strength, as is the use of a real-world pose dataset. However, the empirical support for the headline claim is undermined by two issues: the paper's own stated key metric (AUC-PR) is one on which Shopformer performs worst among the compared methods, and the hyperparameters (token count, embedding size, transformer depth) are selected on the test set, making the reported AUC-ROC optimistic. If these evaluation issues are corrected and the results hold, the method could still be a valuable baseline; as presented, the evidence does not support the claimed state-of-the-art superiority.

major comments (4)
  1. [Sec. 4.2.2 and Table 2] In Section 4.2.2 the paper states that AUC-PR is "a key performance metric for shoplifting detection tasks involving highly imbalanced datasets" and that a higher AUC-PR reflects stronger detection capability. In Table 2, Shopformer's AUC-PR is 44.49%, which is lower than GEPC (50.38%) and far lower than STG-NF (84.06%). On the metric the authors themselves emphasize, Shopformer is not superior; the conclusion in Section 6 that the results "confirm the effectiveness" of the method is therefore not supported by the paper's own evaluation criteria.
  2. [Sec. 4.4 and Table S3] The optimal number of tokens (2), token embedding size (144), and transformer configuration (2 layers, 2 heads, feed-forward dimensionality 64) are selected by comparing AUC-ROC, AUC-PR, and EER on the same test set that is subsequently used for the final comparison in Table 2. Because no held-out validation split is used, the reported 69.15% AUC-ROC is expected to be optimistically biased, and the comparison with baselines, which are run with their default settings, is not an apples-to-apples evaluation.
  3. [Sec. 2 and Sec. 4.3] The paper acknowledges that SPARTA is a relevant transformer-based pose anomaly detection method and an inspiration for the architecture, yet it is excluded from the comparison because its code is not publicly available. Without SPARTA, the claim in Section 6 of outperforming "state-of-the-art pose-based anomaly detection models" omits the closest architectural relative, so the uniqueness and superiority claims remain incomplete.
  4. [Sec. 3.2] The detection mechanism rests on the premise that a transformer trained only on normal pose sequences will reconstruct normal tokens with low MSE and shoplifting tokens with high MSE. This premise is asserted rather than demonstrated. The paper should provide evidence such as the distribution of reconstruction errors for normal versus shoplifting test sequences, or per-incident AUC, to show that the score actually separates the two classes; without such evidence, the empirical results are not interpretable.
minor comments (6)
  1. [Sec. 4.1] The data set name is misspelled as "PeseLift" and should be "PoseLift".
  2. [Sec. 5] The sentence "surpassing the previous state-of-the-art STG-NF model by 67.46% in AUC-ROC" is ambiguous; it should read, for example, "surpassing STG-NF, which achieves 67.46% AUC-ROC" or similar.
  3. [Sec. 3.2] The phrase "compromising L encoder-decoder layers" should be "comprising L encoder-decoder layers."
  4. [Table 1 and Table S2] The main text Table 1 lists AUC-ROC 69.15 for the 2-token, 144-embedding configuration, while the supplementary Table S2 reports 69.14 for the same configuration; the inconsistency should be reconciled.
  5. [Table 2 and Fig. 4] The EER values in Table 2 are rounded to two decimals (0.39, 0.41, 0.38, 0.38), while the text and Figure 4 use 0.3819; a consistent level of precision should be used throughout.
  6. [Table 2] No confidence intervals or standard deviations are reported for the main comparison; at least a few random seeds should be run to assess the stability of the claimed AUC-ROC difference.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the evaluation is empirical and the reported scores are measured, not derived from the model's construction or from a self-citation chain.

full rationale

Shopformer's derivation chain is empirical rather than definitional. The reconstruction-error premise (Sec. 3.2) is a standard unsupervised assumption: train on normal pose sequences, then score test sequences by MSE. The anomaly score is not a fitted parameter renamed as a prediction; it is computed from held-out test inputs after training on normal-only data. The claimed superiority over STG-NF, TSGAD, and GEPC rests on the reported AUC-ROC values in Table 2, which are measurements, not consequences of the model's construction. The PoseLift dataset is authored by the same group [30], and several baselines and architectural inspirations are from the same lab ([24], [25]), but the paper does not invoke a uniqueness theorem, does not define its architecture in terms of its target result, and does not present a fitted value as a prediction. The self-citations supply data, baselines, and design context, not the conclusion that Shopformer wins. Concerns that hyperparameters were selected on the test set (Sec. 4.4, Table S3) and that the AUC-PR results contradict the AUC-ROC headline are methodological or statistical validity issues, not circularity. No equation in the paper reduces to another by construction, and no fitted parameter is relabeled as a prediction. Under the hard rules requiring a quoted reduction, no specific circular step can be exhibited, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

No new physical or conceptual entities are introduced. The model's performance depends on several domain assumptions about pose data and on hyperparameters tuned on the test set, which is the main source of circularity in the evaluation.

free parameters (3)
  • Number of tokens N = 2
    Selected from {1, 2, 3, 4, 6, 12} based on highest AUC-ROC on the test set (Sec 4.4, Table S2), not on a validation split.
  • Token embedding size (GCAE output channels) = 144 (8 channels)
    Selected from {72, 144, 216, 288, 504, 576, 1152} based on test-set AUC-ROC (Table 1 and Table S2).
  • Transformer layers, heads, feed-forward dimension = L=2, T=2, F=64
    Chosen from test-set ablation results in Table S3; differences across configurations are within 0.1 AUC-ROC points, so this is effectively test-set tuning.
assumptions (3)
  • domain assumption Reconstruction error separates normal from shoplifting pose sequences.
    The anomaly score is the MSE between input and reconstructed tokens; Sec 3.2 asserts that shoplifting sequences produce higher MSE, which is an empirical premise rather than a guaranteed property.
  • domain assumption The PoseLift training set contains only normal behaviors and its frame-level labels are accurate.
    Training is unsupervised on normal frames (Sec 4.1); any shoplifting frames in the training set would contaminate the normality model, and label errors would corrupt the evaluation.
  • domain assumption COCO17 2D keypoints from HRNet are sufficient to capture shoplifting behavior.
    The model sees only keypoint coordinates (Sec 3.1 and Sec 4.1), so all behavioral signal must be present in the pose estimates; occlusions or pose errors in a retail environment could break this.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Shopformer: Transformer-Based Framework for Detecting Shoplifting via Human Pose." pith.science (2026). https://pith.science/paper/V7S7LLRM

@misc{pith2026250419970,
  author       = {Pith},
  title        = {Pith review of: Shopformer: Transformer-Based Framework for Detecting Shoplifting via Human Pose},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V7S7LLRM}},
  note         = {Machine review of arXiv:2504.19970}
}
read the original abstract

Shoplifting remains a costly issue for the retail sector, but traditional surveillance systems, which are mostly based on human monitoring, are still largely ineffective, with only about 2% of shoplifters being arrested. Existing AI-based approaches rely on pixel-level video analysis which raises privacy concerns, is sensitive to environmental variations, and demands significant computational resources. To address these limitations, we introduce Shopformer, a novel transformer-based model that detects shoplifting by analyzing pose sequences rather than raw video. We propose a custom tokenization strategy that converts pose sequences into compact embeddings for efficient transformer processing. To the best of our knowledge, this is the first pose-sequence-based transformer model for shoplifting detection. Evaluated on real-world pose data, our method outperforms state-of-the-art anomaly detection models, offering a privacy-preserving, and scalable solution for real-time retail surveillance. The code base for this work is available at https://github.com/TeCSAR-UNCC/Shopformer.

Figures

Figures reproduced from arXiv: 2504.19970 by the authors.

Figure 1
Figure 1. This map shows the revenue lost to retail shoplifting [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Represents the status of facial recognition technology [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the Shopformer architecture. The framework operates in two stages: (1) a Graph Convolutional Autoencoder is [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparison of AUC-ROC, AUC-PR, and EER across [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 32 canonical work pages

  1. [1]

    A survey of anomaly detection in cyber-physical systems

    Danial Abshari and Meera Sridhar. A survey of anomaly detection in cyber-physical systems. arXiv preprint arXiv:2502.13256, 2025. 1

    arXiv 2025

  2. [2]

    Pishgu: Universal path prediction network architecture for real-time cyber-physical edge systems

    Ghazal Alinezhad Noghre, Vinit Katariya, Armin Danesh Pazho, Christopher Neff, and Hamed Tabkhi. Pishgu: Universal path prediction network architecture for real-time cyber-physical edge systems. In Proceedings of the ACM/IEEE 14th International Conference on Cyber- Physical Systems (with CPS-IoT Week 2023) , pages 88–97,

    work page 2023

  3. [3]

    An expert video surveillance system to identify and mitigate shoplifting in megastores

    Mohd Aquib Ansari and Dushyant Kumar Singh. An expert video surveillance system to identify and mitigate shoplifting in megastores. Multimedia Tools and Applications, 81(16): 22497–22525, 2022. 1, 3

    work page 2022

  4. [4]

    Opti- mized parameter tuning in a recurrent learning process for shoplifting activity classification

    Mohd Aquib Ansari and Dushyant Kumar Singh. Opti- mized parameter tuning in a recurrent learning process for shoplifting activity classification. Cybernetics and Informa- tion Technologies, 23(1):141–160, 2023. 1, 3

    work page 2023

  5. [5]

    Understanding ethics, privacy, and regulations in smart video surveillance for public safety

    Babak Rahimi Ardabili, Armin Danesh Pazho, Ghazal Alinezhad Noghre, Christopher Neff, Arun Ravin- dran, and Hamed Tabkhi. Understanding ethics, privacy, and regulations in smart video surveillance for public safety. arXiv preprint arXiv:2212.12936, 2022. 2

    arXiv 2022

  6. [6]

    Understanding policy and technical aspects of ai-enabled smart video surveillance to address public safety

    Babak Rahimi Ardabili, Armin Danesh Pazho, Ghazal Alinezhad Noghre, Christopher Neff, Sai Datta Bhaskararayuni, Arun Ravindran, Shannon Reid, and Hamed Tabkhi. Understanding policy and technical aspects of ai-enabled smart video surveillance to address public safety. Computational Urban Science, 3(1):21, 2023. 2

    work page 2023

  7. [7]

    Exploring public’s perception of safety and video surveillance technology: A survey approach

    Babak Rahimi Ardabili, Armin Danesh Pazho, Ghazal Alinezhad Noghre, Vinit Katariya, Gordon Hull, Shannon Reid, and Hamed Tabkhi. Exploring public’s perception of safety and video surveillance technology: A survey approach. Technology in Society, 78:102641, 2024. 1

    work page 2024

  8. [8]

    Introducing the ai governance and regulatory archive (agora): An analytic infrastructure for navigating the emerging ai governance landscape

    Zachary Arnold, Daniel S Schiff, Kaylyn Jackson Schiff, Brian Love, Jennifer Melot, Neha Singh, Lindsay Jenkins, Ashley Lin, Konstantin Pilz, Ogadinma Enweareazu, et al. Introducing the ai governance and regulatory archive (agora): An analytic infrastructure for navigating the emerging ai governance landscape. In Proceedings of the AAAI/ACM Conference on ...

    work page 2024

Show all 41 references
  1. [9]

    Expert video-surveillance system for real-time detection of suspicious behaviors in shopping malls

    Roberto Arroyo, J Javier Yebes, Luis M Bergasa, Iv ´an G Daza, and Javier Almaz´an. Expert video-surveillance system for real-time detection of suspicious behaviors in shopping malls. Expert systems with Applications, 42(21):7991–8005,

  2. [10]

    Detecting zero- day web attacks using one-class ensemble classifiers

    Vahid Babaey and Hamid Reza Faragardi. Detecting zero- day web attacks using one-class ensemble classifiers. 2025. 1

    2025

  3. [11]

    Deep learning reveals what facial expressions mean to people in different cultures

    Jeffrey A Brooks, Lauren Kim, Michael Opara, Dacher Kelt- ner, Xia Fang, Maria Monroy, Rebecca Corona, Panagiotis Tzirakis, Alice Baird, Jacob Metrick, et al. Deep learning reveals what facial expressions mean to people in different cultures. Iscience, 27(3), 2024. 2

    2024

  4. [12]

    Shoplifting statistics, 2025

    Capital One Shopping. Shoplifting statistics, 2025. Ac- cessed: March 17, 2025. 1

    2025

  5. [13]

    Channel-wise topology refinement graph convolution for skeleton-based action recognition

    Yuxin Chen, Ziqi Zhang, Chunfeng Yuan, Bing Li, Ying Deng, and Weiming Hu. Channel-wise topology refinement graph convolution for skeleton-based action recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 13359–13368, 2021. 2

    2021

  6. [14]

    An adaptive classification system for video-based face recog- nition

    Jean-Franc ¸ois Connolly, Eric Granger, and Robert Sabourin. An adaptive classification system for video-based face recog- nition. Information Sciences, 192:50–70, 2012. 2

    2012

  7. [15]

    An automatic shoplifting detection from surveillance videos (student abstract)

    U-Ju Gim, Jae-Jun Lee, Jeong-Hun Kim, Young-Ho Park, and Aziz Nasridinov. An automatic shoplifting detection from surveillance videos (student abstract). In Proceed- ings of the AAAI Conference on Artificial Intelligence, pages 13795–13796, 2020. 1

    2020

  8. [16]

    Normalizing flows for human pose anomaly detection

    Or Hirschorn and Shai Avidan. Normalizing flows for human pose anomaly detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 13545– 13554, 2023. 3, 6, 8

    2023

  9. [17]

    Rethinking affect analysis: A protocol for ensuring fairness and con- sistency

    Guanyu Hu, Dimitrios Kollias, Eleni Papadopoulou, Paraskevi Tzouveli, Jie Wei, and Xinyu Yang. Rethinking affect analysis: A protocol for ensuring fairness and con- sistency. IEEE Transactions on Biometrics, Behavior, and Identity Science, 2025. 2

    2025

  10. [18]

    Detection of shoplifting on video using a hybrid network

    Lyudmyla Kirichenko, Tamara Radivilova, Bohdan Sydorenko, and Sergiy Yakovlev. Detection of shoplifting on video using a hybrid network. Computation, 10(11):199,

  11. [19]

    Facing facts: The effect of facial recognition bans on policing effectiveness

    Thomas Christopher Latkowski. Facing facts: The effect of facial recognition bans on policing effectiveness. Master’s thesis, Georgetown University, 2024. 2

    2024

  12. [20]

    Graph embedded pose clustering for anomaly detection

    Amir Markovitz, Gilad Sharir, Itamar Friedman, Lihi Zelnik- Manor, and Shai Avidan. Graph embedded pose clustering for anomaly detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 10539–10547, 2020. 2, 3, 4, 6, 8

    2020

  13. [21]

    Face detection without bells and whistles

    Markus Mathias, Rodrigo Benenson, Marco Pedersoli, and Luc Van Gool. Face detection without bells and whistles. In European conference on computer vision, pages 720–735. Springer, 2014. 2

    2014

  14. [22]

    Shoplifting detection using hybrid neural network cnn-bilsmt and development of benchmark dataset

    Iqra Muneer, Mubbashar Saddique, Zulfiqar Habib, and Heba G Mohamed. Shoplifting detection using hybrid neural network cnn-bilsmt and development of benchmark dataset. Applied Sciences, 13(14):8341, 2023. 1, 3

    2023

  15. [23]

    Suspicious behavior detection with temporal feature extraction and time-series classification for shoplifting crime prevention

    Amril Nazir, Rohan Mitra, Hana Sulieman, and Firuz Ka- malov. Suspicious behavior detection with temporal feature extraction and time-series classification for shoplifting crime prevention. Sensors, 23(13):5811, 2023. 1, 3

    2023

  16. [24]

    An exploratory study on human-centric video anomaly detection through variational autoencoders and trajectory prediction

    Ghazal Alinezhad Noghre, Armin Danesh Pazho, and Hamed Tabkhi. An exploratory study on human-centric video anomaly detection through variational autoencoders and trajectory prediction. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages ...

    2024

  17. [25]

    Human-centric video anomaly detection through spatio-temporal pose tokenization and transformer,

    Ghazal Alinezhad Noghre, Armin Danesh Pazho, and Hamed Tabkhi. Human-centric video anomaly detection through spatio-temporal pose tokenization and transformer,

  18. [26]

    Mimicme: A large scale diverse 4d database for facial expression analysis

    Athanasios Papaioannou, Baris Gecer, Shiyang Cheng, Grig- orios Chrysos, Jiankang Deng, Eftychia Fotiadou, Christos Kampouris, Dimitrios Kollias, Stylianos Moschoglou, Kri- taphat Songsri-In, et al. Mimicme: A large scale diverse 4d database for facial expression analysis. In ...

    2022

  19. [27]

    A survey of graph-based deep learning for anomaly detection in distributed systems

    Armin Danesh Pazho, Ghazal Alinezhad Noghre, Arnab A Purkayastha, Jagannadh Vempati, Otto Martin, and Hamed Tabkhi. A survey of graph-based deep learning for anomaly detection in distributed systems. IEEE Transactions on Knowledge and Data Engineering, 36(1):1–20, 2023. 1

    2023

  20. [28]

    Vt-former: An exploratory study on vehicle trajectory prediction for highway surveil- lance through graph isomorphism and transformer

    Armin Danesh Pazho, Ghazal Alinezhad Noghre, Vinit Katariya, and Hamed Tabkhi. Vt-former: An exploratory study on vehicle trajectory prediction for highway surveil- lance through graph isomorphism and transformer. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision ...

    2024

  21. [29]

    Facial recognition technology: regulations, rights and the rule of law

    Mais Qandeel. Facial recognition technology: regulations, rights and the rule of law. Frontiers in big Data, 7:1354659,

  22. [30]

    Exploring pose-based anomaly detection for retail security: A real- world shoplifting dataset and benchmark

    Narges Rashvand, Ghazal Alinezhad Noghre, Armin Danesh Pazho, Shanle Yao, and Hamed Tabkhi. Exploring pose-based anomaly detection for retail security: A real- world shoplifting dataset and benchmark. arXiv preprint arXiv:2501.06591, 2025. 1, 2, 3, 5, 8

    2025 arXiv

  23. [31]

    Trajectron++: Dynamically-feasible tra- jectory forecasting with heterogeneous data

    Tim Salzmann, Boris Ivanovic, Punarjay Chakravarty, and Marco Pavone. Trajectron++: Dynamically-feasible tra- jectory forecasting with heterogeneous data. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XVIII 16, pages...

    2020

  24. [32]

    Real-world anomaly detection in surveillance videos

    Waqas Sultani, Chen Chen, and Mubarak Shah. Real-world anomaly detection in surveillance videos. In Proceedings of the IEEE conference on computer vision and pattern recog- nition, pages 6479–6488, 2018. 3

    2018

  25. [33]

    Census Bureau

    U.S. Census Bureau. Census data, population. https:// data.census.gov/. Accessed: March 20, 2025. 2

    2025

  26. [34]

    Congress

    U.S. Congress. Facial recognition and biometric technology moratorium act of 2023, 2023. Accessed: 2025-03-21. 2

    2023

  27. [35]

    Government Accountability Office

    U.S. Government Accountability Office. Facial recognition technology: Privacy and accuracy issues related to commer- cial uses, 2024. Accessed: 2025-03-21. 2

    2024

  28. [36]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017. 2, 4, 5

    2017

  29. [37]

    Beyond surveillance: privacy, ethics, and regula- tions in face recognition technology

    Xukang Wang, Ying Cheng Wu, Mengjie Zhou, and Hong- peng Fu. Beyond surveillance: privacy, ethics, and regula- tions in face recognition technology. Frontiers in big data, 7:1337465, 2024. 2

    2024

  30. [38]

    Spatial tempo- ral graph convolutional networks for skeleton-based action recognition

    Sijie Yan, Yuanjun Xiong, and Dahua Lin. Spatial tempo- ral graph convolutional networks for skeleton-based action recognition. In Proceedings of the AAAI conference on arti- ficial intelligence, 2018. 2

    2018

  31. [39]

    Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting

    Bing Yu, Haoteng Yin, and Zhanxing Zhu. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. arXiv preprint arXiv:1709.04875 ,

    arXiv

  32. [64]

    11 Table S2

    This setup achieves high accuracy while maintaining a lower parameter count compared to other configurations. 11 Table S2. Performance of Shopformer (AUC-ROC, AUC-PR, EER, EER TH, 10 ER, and 10 ER TH) with token counts of 1, 2, 3, 4, 6, and 12, and token embedding sizes of 72,...

  33. [2017]

    It includes various experimental setups, architectural variations, training strategies, hyperparameter configurations, and detailed results

    2, 4 10 Supplementary Material This section provides a comprehensive overview of the ab- lation studies conducted to determine the optimal configura- tion of Shopformer. It includes various experimental setups, architectural variations, training strategies, hyperparameter conf...

Pith tools

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