Pith. sign in

REVIEW 3 major objections 5 minor 32 references

From Benchmark Performance to Tool Deployment: Human-in-the-Loop Anomaly Detection

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper argues that strong scores on curated anomaly-detection benchmarks do not transfer to a reflective-metal industrial inspection setting, and that robust deployment requires treating preprocessing, nominal-data quality, and human…

desk verdict Real dataset and honest reporting, but the headline swings rest on thresholds chosen from 10 defect images; the paper needs AUROC or bootstrap CIs before the robustness claims are settled. read the letter →

arxiv 2608.07770 v1 pith:E43PIF2S submitted 2026-08-07 cs.LG cs.CV

Mike Szklarzewski , CJ George , Gavin Smithson , Christopher Stokes , Dakota Fulp , William M. Jones , Benjamin Wynn , Alexander Ur
show 5 more authors
Agit Yesiloz Clint Kallenbach Mark Swartz Nathan DeBardeleben Sharmistha Chakrabarti
This is my paper · ORCID
classification cs.LGcs.CV
keywords anomalydetectionindustrialinspectionunsupervisedlearninghuman-in-the-loopbenchmarkevaluationpreprocessingsensitivitydataqualityauditvisual
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 tries to establish that benchmark performance on curated anomaly-detection datasets is not a reliable predictor of deployment behavior in an industrial visual-inspection setting. Evaluating 19 unsupervised anomaly-detection models on the BowTie dataset, a reflective-metal manufacturing dataset with three color profiles and scarce, morphologically varied defects, it finds that no single architecture is uniformly robust: rankings change across profiles, grayscale and resolution preprocessing have strongly architecture-dependent effects, and a consensus audit of nominal training images shows that data quality interacts with model sensitivity. The paper's response is a deployed human-in-the-loop inspection framework that couples annotation, heatmap-guided AI localization, SAM-refined candidate regions, and mask-based validation while keeping final accept/reject authority with the human inspector. If the claim is right, model qualification for such settings must include preprocessing sensitivity and nominal-data auditing rather than a single leaderboard score.

What carries the argument

The load-bearing machinery is a three-part evaluation and audit protocol. First, a profile-wise split trains on nominal images only, uses F1-Max on a small validation set to fix a decision threshold, and reports Macro F1 at that threshold on a held-out test set with very few defect images. Second, controlled preprocessing probes (grayscale conversion and resolution changes) are applied per architecture to measure sensitivity. Third, a consensus-based nominal-data audit reuses the benchmark's own inference outputs to flag training images that six or more models consider anomalous, and the benchmark is rerun without those samples to estimate data-quality effects. The operational counterpart is a human-in-the-loop workflow in which anomaly heatmaps localize candidate defects, the Segment Anything Model (SAM) refines those regions into editable proposals, and inspectors accept, reject, or adjust boundaries before image-level disposition.

What would settle it

Re-run the 19 models on BowTie with many random splits per profile (or a substantially larger collected test set), and check whether the large Macro F1 swings—especially CFlow's 0.170 on CP2 at baseline versus 0.531 after the consensus audit, and its 0.863 on CP1—persist. If the swings shrink to within sampling error, the claim that no architecture is uniformly robust would need to be softened; if they persist, the deployment caution is confirmed.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the BowTie dataset acts as a deployment-oriented stress test exposing fragility that cleaner benchmarks hide. The same 19 models, trained on the same profile-wise splits, produce large Macro F1 swings (CFlow scores 0.863 on CP1 but 0.170 on CP2), preprocessing changes help some architectures and hurt others, and no architecture remains at the top across all three color profiles. A consensus-based audit, in which any nominal training image flagged by six or more of the 19 models is set aside for expert review, shows that rerunning without those images improves some models strongly (CFlow on CP2 rises to 0.531) while degrading others, indicating that nominal-data quality is not a uniform fix. The paper concludes that annotation, inference, and quantitative validation should be treated as connected parts of one inspection workflow, and it presents AnnoMate, MicroSentryAI, and a validation engine as an initially deployed implementation of that principle.

Load-bearing premise

The rankings and sensitivity conclusions rest on a single split per color profile with test sets containing as few as 10 defect images and a threshold chosen from a small validation set, so the large swings could partly reflect sampling noise rather than true model instability.

Editorial extensions

If this is right

  • In this inspection setting, selecting a model by a single benchmark rank is unsafe; candidates should be screened under the exact preprocessing and profile conditions they will face.
  • Grayscale conversion and sensor resolution should be treated as per-model tuning variables, not universal preprocessing defaults.
  • Nominal-data auditing should be part of model qualification, but blanket removal of consensus-flagged images is not a safe default because some models degrade after cleaning.
  • A human-in-the-loop workflow with heatmap review, SAM-refined proposals, and mask-based validation offers a deployment path when automated scores are unstable.
  • The reported rankings are BowTie-specific stress-test results, not general capability claims about the architectures.

Reading between the lines

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

  • If the fragility pattern generalizes, benchmark reporting in industrial anomaly detection should include confidence intervals and condition-specific breakdowns rather than a single macro-F1 number.
  • The consensus audit could be developed into a general pre-deployment data-quality screen: train a diverse panel of models and use cross-family agreement to prioritize which nominal images get expert review.
  • The framework's stored review history and expert-comparison mode set up a natural before/after study of inspector consistency and onboarding time, which the paper lists as future work.
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

3 major / 5 minor

Summary. The paper evaluates 19 unsupervised anomaly-detection models from the Anomalib library on the BowTie dataset, a real manufacturing dataset with reflective surfaces, subtle defects, and three color profiles. It reports that no model is uniformly robust across profiles, that grayscale and resolution preprocessing have strongly architecture-dependent effects, and that a consensus-based audit of nominal training images suggests that nominal-data quality interacts with model sensitivity. On the basis of these findings, the paper presents a human-in-the-loop inspection framework (AnnoMate, MicroSentryAI, and a Validation Engine) that was initially deployed in a manufacturing workflow. The central claim is that benchmark performance, as typically reported on curated datasets such as MVTec AD, is not a reliable guide to deployment in this industrial inspection setting.

Significance. If the empirical conclusions are supported, the paper makes a useful contribution to the growing literature on the benchmark-to-deployment gap in industrial anomaly detection. The BowTie dataset is a realistic stress test with reflective metal surfaces and profile-specific appearance shift, and the 19-model comparison under a common framework is valuable. The paper is transparent about several limitations, provides a reproducibility link, and describes a deployed human-in-the-loop tool that directly addresses operational needs. The main significance lies in the negative result that model rankings and preprocessing defaults do not transfer across regimes, and in the proposed workflow that couples annotation, inference, and validation. However, the strength of this significance depends on whether the measured instability is real or partly an artifact of the evaluation protocol.

major comments (3)
  1. [Section III-C, Tables IV-VI] The paper's central claim that 'no single architecture emerged as uniformly robust' and that preprocessing effects are strongly architecture-dependent rests on point estimates of Macro F1 at a single F1-Max threshold selected on the validation split. For CP2, the validation and test sets each contain only 10 defect images (Table III), so threshold selection is extremely noisy. The CFlow CP2 result in Table IV (Macro F1 = 0.170, anomalous-class F1 = 0.339) implies a nominal-class F1 of approximately 0.001, i.e., the chosen threshold classifies essentially every nominal test image as anomalous; this is a degenerate operating point that likely reflects threshold-selection noise rather than a stable property of the model. The large swings used to argue for instability (e.g., CFlow 0.863 on CP1 vs 0.170 on CP2, and the post-cleaning 0.531 in Table VI) are therefore not interpretable without confidence intervals, bootstrap replicates, or a threshold-free metric such as image-level AUROC. Please add uncertainty quantification, and consider reporting AUROC or average precision alongside the F1-Max-based numbers.
  2. [Sections III-D and V] The consensus-based data audit is circular with respect to the claim that nominal-data quality affects deployment. The audit flags training images using the same 19 models whose performance is then reported after those flagged samples are removed (Table VI). The paper explicitly acknowledges this limitation and calls the reruns a 'discussion-oriented data-quality analysis,' but the Conclusion nevertheless states that the audit 'indicated that nominal-data quality further interacts with model sensitivity.' This inference requires an independent check of the flagged samples, for example, a blind SME review of a random sample of consensus-flagged images, or an evaluation on test images that were not used to derive the consensus flag. Without such a check, the post-audit improvements could reflect the models' own training-set contamination or threshold shifts rather than genuine label noise in the nominal data. Please either provide independent validation of the flagged subset or restrict the claim to 'the audit flags candidates for SME review' without asserting a causal quality interaction.
  3. [Section IV-B, Table V] The preprocessing sensitivity analysis is reported without uncertainty measures, and several entries are missing or non-comparable across architectures. The notation '±.000' is unexplained, dashes denote omitted ablations, and U-Flow and Dinomaly already operate at 448×448 at baseline, so the resolution sweep is not uniform. The deltas are computed from single runs; for example, CFlow CP2 +0.558 under grayscale has the same magnitude as the baseline swing that the paper interprets as instability, so without error bars the claim that 'preprocessing choices had strongly architecture-dependent effects' is not yet distinguishable from measurement noise. Please provide replicated runs or at least bootstrap confidence intervals for the deltas, and state explicitly how missing cells and baseline-resolution differences are handled when interpreting Table V.
minor comments (5)
  1. [Abstract and Introduction] There is a typo in the Introduction: 'theBowTiedataset' should read 'the BowTie dataset'.
  2. [Table I] The 'Totals' row reports 1241 good, 119 reject, and 1360 total, but the 'Used' column sums to 1130 good and 94 defect; adding a 'Used subtotal' row would make the relationship between the raw acquisition set and the benchmark subset clearer.
  3. [Section III-C] The phrase 'F1-Max denotes the highest F1 score achievable by sweeping all possible decision thresholds' is somewhat imprecise because F1-Max is not a single threshold; consider rewording to 'the threshold is selected by maximizing the validation-set F1 score.'
  4. [Figures 3 and 8] Figure 3 is referenced after Figure 8 in the text, and Figures 7 and 8 use MVTec examples; the captions should explicitly note that the IoU values shown are on MVTec data, not on BowTie, to avoid confusion.
  5. [Reference [33]] The reproducibility link uses a placeholder URL ('tinyurl.com/placeholder'); this must be replaced with the actual repository location before publication.

Circularity Check

1 steps flagged · score 4.0 of 10

Consensus data audit is self-referential: the same models define the flagged subset and then show post-cleaning gains, so the audit-backed conclusion is partially circular; the primary benchmark itself is independent.

  1. other [Section III-D and Section V, Table VI]
    "We then aggregated the model outputs and marked any training image flagged as anomalous by six or more benchmark models as a consensus-flagged nominal sample. ... The most striking individual result is CFlow, which recovers from very low baseline performance (Macro F1 = 0.170) to 0.531 after cleaning—a result to interpret cautiously given the small CP2 test support, but one that illustrates why nominal-data auditing is especially consequential for likelihood-based architectures."

    The 'consensus-flagged nominal subset' is defined by the anomaly scores of the benchmark models themselves, and the post-audit improvements are measured on those same models after the flagged training samples are removed. This makes the audit a self-referential signal rather than an independent measure of nominal-data quality: the samples removed are, by construction, the ones the models already consider most inconsistent with the learned nominal support. Reporting that those models improve after removing those samples does not independently establish that nominal-data quality caused the improvement, since the selection rule and the outcome are generated by the same models.

full rationale

The primary benchmark study (Sections III-C and IV) is self-contained with respect to circularity: models are trained on nominal-only splits, validation thresholds are chosen by F1-Max, and test Macro F1 is reported on held-out images without test labels entering training or threshold selection. The central claim that no single architecture is uniformly robust rests on these external comparisons and is not derived from its own inputs. The one genuinely circular element is the consensus-based data audit (Section III-D and Table VI): the flagged subset is built from the benchmark models' own training-set anomaly scores, and the same models' post-cleaning gains are then presented as evidence that nominal-data quality interacts with model sensitivity. The paper's caveats reduce the severity, but the abstract and conclusion still rely on the audit as supporting the main thesis, warranting a score of 4 rather than 0. Threshold-selection instability on small CP2 validation support is a statistical validity concern, not a circularity, so it does not raise the circularity score.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The only hand-chosen number that materially affects a secondary result is the six-model consensus threshold. The main benchmark relies on standard domain assumptions about Anomalib fidelity, image-level evaluation, and dataset realism; no new physical or abstract entities are postulated.

free parameters (1)
  • Consensus flagging threshold = 6 models
    Chosen by hand in Section III-D to require cross-family agreement while keeping the audit conservative for SME triage; it is an ad hoc cutoff, not derived from a separate criterion.
assumptions (4)
  • domain assumption Anomalib implementations faithfully represent the published algorithms of the 19 models.
    Section III-C states all experiments used Anomalib and standardized splits and early stopping; the paper assumes implementation fidelity.
  • domain assumption Image-level F1 at a threshold selected on validation is a sufficient benchmark signal without pixel masks.
    Section III-C notes BowTie has no pixel-level masks, so model and threshold selection are done at image level; this is an explicit scope limitation.
  • domain assumption The BowTie dataset (as filtered) is a meaningful proxy for real-world deployment stress.
    Section III-A frames BowTie as a deployment-oriented stress test; the paper relies on the dataset's realism to generalize the conclusions.
  • ad hoc to paper A six-model consensus flag is a valid indicator of nominal-data contamination.
    Section III-D introduces the six-model rule solely for the audit; it is not validated against external labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Benchmark Performance to Tool Deployment: Human-in-the-Loop Anomaly Detection." pith.science (2026). https://pith.science/paper/E43PIF2S

@misc{pith2026260807770,
  author       = {Pith},
  title        = {Pith review of: From Benchmark Performance to Tool Deployment: Human-in-the-Loop Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E43PIF2S}},
  note         = {Machine review of arXiv:2608.07770}
}
read the original abstract

Automated anomaly detection methods often report strong performance on curated academic benchmarks, but their behavior under real-world industrial conditions is less clear. In this work, we evaluate 19 unsupervised anomaly detection models on the BowTie dataset, a challenging manufacturing dataset with reflective surfaces, subtle defects, and profile-specific variation. In contrast to benchmark results, we observe that model performance is less stable than typically reported on standard benchmarks such as MVTec AD, highly sensitive to preprocessing, and inconsistent across conditions, with no single approach emerging as uniformly robust; a consensus audit further indicates that nominal-data quality affects deployment. Motivated by these findings, we developed and initially deployed a unified human-in-the-loop framework for manufactured-part inspection that combines image annotation, AI-assisted defect detection, and an integrated validation engine, replacing a prior manual visual inspection and documentation workflow. The system supports heatmap-guided defect review, SAM-refined candidate regions for inspector acceptance, rejection, or boundary adjustment, mask evaluation where annotations exist, and review history for inspector consistency and onboarding. Together, the results highlight the gap between benchmark performance and deployment reality, and provide a practical framework for addressing it.

Figures

Figures reproduced from arXiv: 2608.07770 by the authors.

Figure 1
Figure 1. Representative tray images for CP1–CP3, with ex [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Examples of reject-class defects across CP1-CP3. Red [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. EfficientAD detection results on four CP1 (left) and four CP3 (right) defect samples: originals, heatmaps, overlaid [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: AnnoMate & MicroSentryAI human-in-the-loop life [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The AnnoMate interface for defect annotation. It sup￾ports polygon-based labeling, metadata tracking, and dataset review to streamline ground-truth curation. and Anomalib-compatible outputs. Inspectors can review pre￾dictions, tune inference parameters, and return acce…
Figure 7
Figure 7. Figure 7: Mask-overlap visualization evaluating a PatchCore [ [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: The validation interface for quantitative mask evalu [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 15 canonical work pages

  1. [1]

    Oceans 11 Data Repository

    Los Alamos National Laboratory, “Oceans 11 Data Repository.” https: //oceans11.lanl.gov, 2026. Accessed: Apr. 13, 2026

    work page 2026

  2. [2]

    Annomate with microsentryai: Human-in-the-loop aiml-powered inspection tools

    Los Alamos National Laboratory, “Annomate with microsentryai: Human-in-the-loop aiml-powered inspection tools.” https://github.com/ lanl/annomate-microsentryai-workflow, 2026. GitHub repository, ac- cessed April 2026

    work page 2026

  3. [3]

    Anomaly detection in laser powder bed fusion using machine learning: A review,

    T. Saharet al., “Anomaly detection in laser powder bed fusion using machine learning: A review,”Results Eng. doi: 10.1016/j.rineng.2022.100803, vol. 17, 2023

    arXiv 2022

  4. [4]

    Automatic metal parts inspection: Use of thermographic images and anomaly detection algorithms,

    M. Benmoussat, M. Guillaume, Y . Caulier, and K. Spinnler, “Automatic metal parts inspection: Use of thermographic images and anomaly detection algorithms,”Infrared Phys. & Tech. doi: 10.1016/j.infrared.2013.07.007, vol. 61, pp. 68–80, 2013

    work page doi:10.1016/j.infrared.2013.07.007 2013

  5. [5]

    A deep-learning-based in-situ surface anomaly detection methodology for laser directed energy deposition via powder feeding,

    F. Kajiet al., “A deep-learning-based in-situ surface anomaly detection methodology for laser directed energy deposition via powder feeding,” J. Manuf. Process. doi: 10.1016/j.jmapro.2022.06.046, vol. 81, pp. 624– 637, 2022

    work page doi:10.1016/j.jmapro.2022.06.046 2022

  6. [6]

    The via annotation software for images, audio and video,

    A. Dutta and A. Zisserman, “The via annotation software for images, audio and video,” inACM Multimedia doi: 10.1145/3343031.3350535, 2019

    arXiv 2019

  7. [7]

    Labelme: A database and web-based tool for image annotation,

    B. C. Russell, A. Torralba, K. P. Murphy, and W. T. Freeman, “Labelme: A database and web-based tool for image annotation,”Int. J. Comput. Vis. doi: 10.1007/s11263-007-0090-8, vol. 77, no. 1–3, pp. 157–173, 2008

    doi:10.1007/s11263-007-0090-8 2008

  8. [8]

    Anomaly detection: A survey,

    V . Chandola, A. Banerjee, and V . Kumar, “Anomaly detection: A survey,” ACM Comput. Surv., vol. 41, July 2009. doi: 10.1145/1541880.1541882

    arXiv 2009

Show all 32 references
  1. [9]

    Deep Learning for Anomaly Detection: A Review doi: 10.1145/3439950,

    G. Pang, C. Shen, L. Cao, and A. V . D. Hengel, “Deep Learning for Anomaly Detection: A Review doi: 10.1145/3439950,”ACM Comput. Surv., Mar. 2021

    2021 doi

  2. [10]

    Anomalib: A Deep Learning Library for Anomaly Detection doi: 10.1109/ICIP46576.2022.9897283,

    S. Akcayet al., “Anomalib: A Deep Learning Library for Anomaly Detection doi: 10.1109/ICIP46576.2022.9897283,” inICIP, pp. 1706– 1710, 2022

    2022

  3. [11]

    What makes a good data augmentation for few-shot unsupervised image anomaly detection?,

    L. Zhanget al., “What makes a good data augmentation for few-shot unsupervised image anomaly detection?,” 2023. arXiv:2304.03294

    2023 arXiv

  4. [12]

    The MVTec anomaly detection dataset: A comprehensive real-world dataset for unsupervised anomaly detection,

    P. Bergmann, K. Batzner, M. Fauser, D. Sattlegger, and C. Steger, “The MVTec anomaly detection dataset: A comprehensive real-world dataset for unsupervised anomaly detection,”Int. J. Comput. Vis., vol. 129, no. 4, pp. 1038–1059, 2021. doi: 10.1007/s11263-020-01400-4

    2021 doi

  5. [13]

    Deep industrial image anomaly detection: A survey,

    J. Liuet al., “Deep industrial image anomaly detection: A survey,”Mach. Intell. Res., vol. 21, no. 1, pp. 104–135, 2024. doi: 10.1007/s11633-023- 1459-z

    2024 doi

  6. [14]

    IM-IAD: Industrial image anomaly detection benchmark in manufacturing,

    G. Xieet al., “IM-IAD: Industrial image anomaly detection benchmark in manufacturing,” inCVPR, 2024. arXiv:2301.13359

    2024 arXiv

  7. [15]

    EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies,

    K. Batzner, L. Heckler, and R. Konig, “EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies,” inWACV, pp. 127– 137, 2024. doi: 10.1109/W ACV57701.2024.00020

    2024

  8. [17]

    Student-teacher feature pyramid matching for anomaly detection,

    G. Wang, S. Han, E. Ding, and D. Huang, “Student-teacher feature pyramid matching for anomaly detection,” inBMVC, 2021. doi: 10.5244/C.35.349

    2021 doi

  9. [18]

    Towards Total Recall in Industrial Anomaly Detection,

    K. Roth, L. Pemula, J. Zepeda, B. Scholkopf, T. Brox, and P. Gehler, “Towards Total Recall in Industrial Anomaly Detection,” inCVPR, pp. 14318–14328, 2022. doi: 10.1109/CVPR52688.2022.01392

    2022

  10. [19]

    PaDiM: A patch dis- tribution modeling framework for anomaly detection and localization,

    T. Defard, A. Setkov, A. Loesch, and R. Audigier, “PaDiM: A patch dis- tribution modeling framework for anomaly detection and localization,” inICPR Workshops, pp. 475–489, 2021. doi: 10.1007/978-3-030-68799- 1 35

    2021 doi

  11. [20]

    CFA: Coupled-hypersphere-based fea- ture adaptation for target-oriented anomaly localization,

    S. Lee, S. Lee, and B. C. Song, “CFA: Coupled-hypersphere-based fea- ture adaptation for target-oriented anomaly localization,”IEEE Access, vol. 10, pp. 78446–78454, 2022. doi: 10.1109/ACCESS.2022.3193699

    2022

  12. [21]

    AnomalyDINO: Boosting Patch-based Few-Shot Anomaly Detection with DINOv2,

    S. Damm, M. Laszkiewicz, J. Lederer, and A. Fischer, “AnomalyDINO: Boosting Patch-based Few-Shot Anomaly Detection with DINOv2,” in WACV, pp. 1319–1329, 2025. doi: 10.1109/W ACV61041.2025.00136

    2025

  13. [22]

    SuperSimpleNet: Unifying Unsu- pervised and Supervised Learning for Fast and Reliable Surface Defect Detection,

    B. Rolih, M. Fu ˇcka, and D. Sko ˇcaj, “SuperSimpleNet: Unifying Unsu- pervised and Supervised Learning for Fast and Reliable Surface Defect Detection,” inPattern Recognit., pp. 47–65, 2025. arXiv:2408.03143

    2025 arXiv

  14. [23]

    FastFlow: Unsupervised Anomaly Detection and Localization via 2D Normalizing Flows,

    J. Yu, Y . Zheng, X. Wang, W. Li, Y . Wu, R. Zhao, and L. Wu, “FastFlow: Unsupervised Anomaly Detection and Localization via 2D Normalizing Flows,” 2021. arXiv:2111.07677

    2021 arXiv

  15. [25]

    Fully Convolu- tional Cross-Scale-Flows for Image-based Defect Detection,

    M. Rudolph, T. Wehrbein, B. Rosenhahn, and B. Wandt, “Fully Convolu- tional Cross-Scale-Flows for Image-based Defect Detection,” inWACV, pp. 1829–1838, 2022. doi: 10.1109/W ACV51458.2022.00189

    2022

  16. [26]

    U-Flow: A u-shaped normalizing flow for anomaly detection with unsupervised threshold,

    M. Tailanian, ´A. Pardo, and P. Mus ´e, “U-Flow: A u-shaped normalizing flow for anomaly detection with unsupervised threshold,”J. Math. Imaging Vis., vol. 66, no. 4, pp. 678–696, 2024. doi: 10.1007/s10851- 024-01193-y

    2024 doi

  17. [27]

    Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection,

    J. Guoet al., “Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection,” inCVPR, pp. 20405–20415, 2025. paper: CVPR OA

    2025

  18. [28]

    DRÆM – A discriminatively trained reconstruction embedding for surface anomaly detection,

    V . Zavrtanik, M. Kristan, and D. Skocaj, “DRÆM – A discriminatively trained reconstruction embedding for surface anomaly detection,” in ICCV, pp. 8310–8319, 2021. doi: 10.1109/ICCV48922.2021.00822

    2021

  19. [29]

    DSR – a dual subspace re- projection network for surface anomaly detection,

    V . Zavrtanik, M. Kristan, and D. Sko ˇcaj, “DSR – a dual subspace re- projection network for surface anomaly detection,” inECCV, pp. 539– 554, 2022. doi: 10.1007/978-3-031-19821-2 31

    2022 doi

  20. [30]

    GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training [arXiv],

    S. Akcay, A. Atapour-Abarghouei, and T. P. Breckon, “GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training [arXiv],” inACCV, 2018

    2018

  21. [31]

    FRE: A Fast Method For Anomaly Detection And Segmentation [arXiv],

    I. Ndiour, N. Ahuja, U. Genc, and O. Tickoo, “FRE: A Fast Method For Anomaly Detection And Segmentation [arXiv],” 2022

    2022

  22. [32]

    Probabilistic Modeling of Deep Features for Out-of-Distribution and Adversarial Detection,

    N. A. Ahuja, I. Ndiour, T. Kalyanpur, and O. Tickoo, “Probabilistic Modeling of Deep Features for Out-of-Distribution and Adversarial Detection,” 2019. arXiv:1909.11786

    2019 arXiv

  23. [33]

    Repository for reproducibility for ICMLA 2026

    Los Alamos National Laboratory, “Repository for reproducibility for ICMLA 2026.” https://tinyurl.com/placeholder, 2026. Accessed: 2026

    2026

  24. [34]

    SAM 2: Segment Anything in Images and Videos,

    N. Raviet al., “SAM 2: Segment Anything in Images and Videos,” in ICLR, 2025. paper: OpenReview

    2025

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.