Pith. sign in

REVIEW 4 major objections 5 minor 72 references

TMUAD: Enhancing Logical Capabilities in Unified Anomaly Detection Models with a Text Memory Bank

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A text memory bank turns object counts, positions, and sizes into logical anomaly detectors.

desk verdict Class-level text memory bank for logical anomaly detection gives a real, well-ablated gain (UniVAD + MT: +9.3 AUROC on MVTec LOCO), but the strict string matching of VLM positions needs repeatability evidence and the hyperparameters are tuned without a validation protocol. read the letter →

arxiv 2508.21795 v1 pith:ZMYQXYUQ submitted 2025-08-29 cs.CV cs.AI

Jiawei Liu , Jiahe Hou , Wei Wang , Jinsong Du , Yang Cong , Huijie Fan This is my paper
classification cs.CVcs.AI
keywords anomalydetectionlogicalstructuralmemorybankvision-languagemodeltextdescriptionunifiedMVTecLOCO
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

TMUAD makes a specific bet: the logical relationships that anomaly detection needs—which objects are present, how many, where they sit relative to the frame, and how large they are—can be written down as short text strings and stored in a memory bank of normal descriptions, and logical anomalies can then be flagged by comparing the query's text tuple against that memory. The paper builds this class-level text memory bank alongside an object-level image memory bank (segmented objects, preserving contours) and a patch-level image memory bank (CLIP + DINOv2 features) for structural anomalies. On MVTec LOCO, the text bank alone reaches 88.4 image-level AUROC for logical anomalies, and adding it to UniVAD lifts the unified score from 81.9 to 91.2. Across seven industrial and medical datasets, the full three-bank framework reports state-of-the-art performance in the comparisons shown. The insight matters because previous unified methods try to coax logical relationships out of high-dimensional image features, most of which are irrelevant to logic; text is a compressed, explicit representation of exactly the information logic requires.

What carries the argument

The carrier of the argument is the class-level text memory bank MT, together with the logic-aware text extractor and the strict matching rule in Algorithm 2. For each training image, MT stores per-category tuples {class, count, fuzzy position, size}; at inference, Algorithm 2 compares the query's tuples to the most similar normal text and treats strict inequality in count or position, over- or under-sized objects, or missing categories as anomalies, using the corresponding object masks for pixel-level scores. This text-based symbolic comparison replaces graph or similarity-matrix constructions over image features as the logical-anomaly mechanism.

What would settle it

Take a set of normal images from one MVTec LOCO category, run the logic-aware text extractor several times with perturbed crops or lighting and with repeated VLM calls, and count how often Algorithm 2 flags an object-position or count anomaly on normal images. If the false-positive rate from VLM phrasing variance is comparable to the true logical anomaly detection rate, the strict string comparison cannot bear the reported logical AUROC.

Watch

Extended reading notes

Core claim

The central claim is that a class-level text memory bank—structured descriptions per object category: class, count, fuzzy position string, and size percentage—performs logical anomaly detection better than image-feature logical modeling, and that combining this text bank with two image memory banks yields a single framework that handles both logical and structural anomalies. The text tuples are extracted by segmenting objects (Grounded SAM), naming them (CLIP), asking a VLM for fuzzy positions, and computing size percentages. At test time, the query tuple is matched against the nearest normal description; mismatched count or position, size outside the normal range, or a category absent from

Load-bearing premise

The VLM produces the same position phrase for the same visual layout every time, so that a strict string inequality between query and memory really means a positional anomaly rather than a rephrasing of the same arrangement.

Editorial extensions

If this is right

  • Any structural-anomaly detector can gain logical ability by attaching MT: UniVAD jumps from 81.9 to 91.2 AUROC on MVTec LOCO, and PBAS+MT reaches 92.0.
  • Logical anomalies can be localized at pixel level, not just flagged per image, because mismatched text fields point back to the object masks that caused the mismatch.
  • The same three-bank recipe transfers across industrial and medical images, since the text extractor runs on categories defined per dataset rather than fixed defect types.
  • The strict-matching design means performance depends on the diversity of normal text tuples; with enough normal samples the VLM position strings saturate the vocabulary, which is why full-data results exceed few-shot results.
  • Even without the large VLM, a reduced variant (TMUAD-U-S) holds 94.1 AUROC on MVTec LOCO, so the text-bank gain is not simply an artifact of the VLM's scale.

Reading between the lines

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

  • The strict string-equality rule likely underestimates logical AUROC under VLM phrasing variance; a paraphrase-tolerant comparator or normalized position vocabulary should raise the ceiling and stabilize few-shot performance.
  • Because MT stores explicit text, it yields human-readable reasons for each anomaly (for example, 'container moved from Center-Center to Center-Left'), making auditability feasible on top of detection.
  • The same text-bank idea could extend to reconstruction-based or diffusion detectors, which are currently weak on logical anomalies, by supplying a symbolic consistency check independent of image reconstruction.
  • The fuzzy-position vocabulary may drift with image resolution or domain shift, so position strings should be recalibrated per dataset; the paper's per-dataset CLIP fine-tuning suggests such calibration helps.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. TMUAD proposes a three-memory-bank framework for unified logical and structural anomaly detection. It constructs a class-level text memory bank (MT) by using Grounded SAM, CLIP, and Qwen2-VL to produce structured textual descriptions of object category, count, fuzzy position, and size; an object-level image memory bank (MO) from segmented objects via SAM and CLIP; and a patch-level image memory bank (MP) from CLIP/DINOv2 features. At test time, each memory bank produces an anomaly score, and the three scores are fused with fixed weights (Eq. 5). The paper claims state-of-the-art image-level AUROC on MVTec LOCO (94.6) and strong results on six industrial/medical structural datasets. A central plug-and-play result is that adding MT to UniVAD raises MVTec LOCO image-level AUROC from 81.9 to 91.2 (Table VII).

Significance. If the claims hold, the paper makes a useful contribution: it shows that structured textual descriptions of object categories, counts, positions, and sizes can serve as a sufficient carrier for logical anomaly scoring, and that a text memory bank can be plugged into existing detection frameworks (Table VII, column 10 vs column 8). The three-memory-bank ablations in Tables V-VII are informative, and the limitation section is candid about failure cases and few-shot degradation. The main risks are reproducibility and robustness: the logical anomaly score depends on an undefined search procedure and on strict string comparisons of VLM outputs whose repeatability is not analyzed. These issues are fixable within the manuscript's scope.

major comments (4)
  1. [Algorithm 2, line 3; Section III-C1] The function `Search(T_q_img, MT)` is never defined. It is the first step of the class-level text anomaly score and determines which normal text Tsim is compared against the query. The reader cannot tell whether Tsim is the single most similar training description in an embedding space (CLIP text? T5??) or a per-category nearest neighbor, nor which similarity metric is used. Table VI, row 5, reports the ``Class-level Text Matching only MT'' result (88.4 AUROC) without specifying the representation used for selecting Tsim. This is load-bearing for the logical AUROC and for the UniVAD+MT gain in Table VII; please specify the algorithm and ideally ablate the choice of retrieval representation.
  2. [Algorithm 2, lines 12-13; Section V-B] The logical anomaly test treats `O_j_pos != O_t_pos` as an anomaly signal. Position strings are produced once by Qwen2-VL (Algorithm 1, line 15) with no normalization, tolerance, or repeated sampling. VLM outputs are not deterministic; phrasing variants such as ``Center-Center'' versus ``center'' on two normal images with the same layout will cause every normal query to be scored as a position anomaly. The paper's own few-shot analysis (Section V-B) states that the strict text matching strategy is ``overly restrictive'' with limited data; the same fragility can appear at full data if VLM phrasing varies. Please provide a repeatability study of the VLM position outputs on normal images, or introduce a fuzzy/soft matching rule, and quantify the effect on the reported logical AUROC in Table VII.
  3. [Section III-C4, Eq. (5); Section IV-B] The fusion weights lambda_1=0.05, lambda_2=0.3, lambda_3=0.65 are set empirically, but the paper reports no validation split, sensitivity analysis, or selection procedure. Because the method is unsupervised and the test labels are not used for training, it is essential to know whether these weights were selected on the test set; if so, the headline 94.6 AUROC may be optimistically biased. Similarly, the K-means cluster counts (100 for MP, 1000 for MO) and the CLIP fine-tuning sample count (200 per category) are given without a validation protocol. Please add a sensitivity study for lambda_i and clarify how all hyperparameters were chosen.
  4. [Algorithm 1, lines 7-14 and Algorithm 2, line 22] The scope of `Ic` is inconsistent. In Algorithm 1, line 7 initializes `{Ic | c in C} <- 0` inside the per-image loop, so `Ic` accumulates masks only within one training image. However, Algorithm 2, line 22 uses `Ic` as ``all possible regions where the categories Ot_class could occur,'' which suggests a global mask accumulated over the whole training set. The pseudocode does not explain where such a global mask is stored or how query-image masks are aligned with training-image masks. Since missing-category detection is a logical signal, this ambiguity prevents exact reproduction of the ST computation. Please clarify the definition and storage of Ic.
minor comments (5)
  1. [Table I vs Table IV] The reported MVTec LOCO image-level AUROC for Ours is 94.6 in Table I but 95.2 in Table IV's ``Average'' column. If these numbers are computed on different aggregations (per-category mean vs mean of logical and structural subaverages), please state this explicitly to avoid an apparent inconsistency.
  2. [Eq. (1)-(2), Algorithm 2 lines 14-18] The size deviation coefficient alpha divides by (s_max - s_min). If all normal images in a category have the same object size, this denominator is zero. Please add a guard or specify the behavior in that degenerate case.
  3. [Algorithm 1 and Section III-B1] The notation `MT := {T_i_img}ni_i=1` is confusing; the index i already denotes the normal image number, and the lower/upper bound should be typeset as `{T_img^i}_{i=1}^{n_i}`. The current notation appears to place `ni` in the subscript in two roles.
  4. [Tables II and III] The column header ``PBAS+MT / PBAS'' is ambiguous. Please use separate columns or a clear note so the reader knows which numbers correspond to PBAS+MT and which to PBAS alone.
  5. [Algorithm 2, line 21] The condition `if Ot_class not in {O_j_class}` uses a subscript `t` from the retrieved Tsim, while the set notation uses `O_j_class`; this is typeset inconsistently. Please unify the index notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: TMUAD is an empirical memory-bank retrieval method; its scores are not equivalent to its inputs by construction.

full rationale

TMUAD is an empirical retrieval-and-matching method, not a derivation from first principles. The class-level text memory bank MT is constructed by running the logic-aware text extractor (Grounded SAM + CLIP + Qwen2-VL + area calculation) on normal training images; query images undergo the same extraction and are scored by nearest-neighbor retrieval against MT (Algorithm 2). This is a standard memory-bank k-NN formulation: the anomaly score compares a query representation to stored normal representations, and no fitted parameter is renamed as a prediction. The fusion weights (λ1=0.05, λ2=0.3, λ3=0.65), cluster counts, and strict string-matching rule are empirically chosen decision rules; tuning them on the benchmark is a potential evaluation-leakage concern, but it does not make the reported AUROC equivalent to the inputs by construction. The paper's own limitations (Sec. V-A) and few-shot discussion (Sec. V-B) explicitly acknowledge that strict text matching can be overly restrictive, which further indicates the score is not tautologically correct. The only self-citation is reference [13] (same first author), used as a background diffusion example and not load-bearing. The reader's concern about VLM position-string variance (Algorithm 2, lines 12-13) is a robustness/validity threat, not circularity: output noise would degrade performance, but the score is not definitionally tied to the target label. No circular step can be exhibited, so the score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method relies on off-the-shelf foundation models (Grounded SAM, SAM, CLIP, DINOv2, Qwen2-VL) and on normal-only training data. No new physical or ontological entities are introduced; the three memory banks are computational structures built from existing components. The main paid-for inputs are the correctness of these pretrained components, the stability of VLM position descriptions under exact string comparison, and the hand-selected hyperparameters.

free parameters (4)
  • Fusion weights lambda_1, lambda_2, lambda_3 = 0.05, 0.30, 0.65
    Set empirically in Sec. III-C4 (Eq. 5) without a reported validation split; directly controls the final anomaly score and plausibly tuned on MVTec LOCO.
  • K-means cluster count for patch-level memory bank M_P = 100 per image category
    Chosen in Sec. IV-B; Fig. 18 ablates cluster count using AUROC, indicating selection on the evaluation metric.
  • K-means cluster count for object-level memory bank M_O = 1000 per object category
    Chosen in Sec. IV-B; no ablation or justification for this particular value.
  • CLIP fine-tuning sample count = 200 images per category
    Chosen in Sec. IV-B to fine-tune CLIP for the text extractor; no sensitivity analysis is reported.
assumptions (5)
  • domain assumption Grounded SAM and CLIP correctly segment and classify objects in normal images.
    Invoked in Sec. III-B1; if segmentation or classification is wrong, the text memory bank encodes incorrect categories, counts, and sizes.
  • domain assumption Qwen2-VL fuzzy position strings are semantically consistent and comparable by exact string equality.
    Algorithm 2 lines 12-13 treat position mismatch as an anomaly; VLM outputs such as 'Center-Center' may vary across images, causing false positives.
  • domain assumption Pre-trained CLIP and DINOv2 patch features capture structural anomalies through cosine similarity.
    Used in Sec. III-B3 and III-C3; this is the standard PatchCore-style assumption.
  • domain assumption Memory banks built from normal training data are representative and free of anomalies.
    Standard unsupervised AD assumption stated in Sec. IV-A; contamination of training data would degrade the banks.
  • domain assumption K-means clustering preserves the nearest-neighbor retrieval behavior of the full memory bank.
    Used in Sec. III-B2/3 and IV-B; clustering changes similarity ranking and can affect results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TMUAD: Enhancing Logical Capabilities in Unified Anomaly Detection Models with a Text Memory Bank." pith.science (2026). https://pith.science/paper/ZMYQXYUQ

@misc{pith2026250821795,
  author       = {Pith},
  title        = {Pith review of: TMUAD: Enhancing Logical Capabilities in Unified Anomaly Detection Models with a Text Memory Bank},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZMYQXYUQ}},
  note         = {Machine review of arXiv:2508.21795}
}
read the original abstract

Anomaly detection, which aims to identify anomalies deviating from normal patterns, is challenging due to the limited amount of normal data available. Unlike most existing unified methods that rely on carefully designed image feature extractors and memory banks to capture logical relationships between objects, we introduce a text memory bank to enhance the detection of logical anomalies. Specifically, we propose a Three-Memory framework for Unified structural and logical Anomaly Detection (TMUAD). First, we build a class-level text memory bank for logical anomaly detection by the proposed logic-aware text extractor, which can capture rich logical descriptions of objects from input images. Second, we construct an object-level image memory bank that preserves complete object contours by extracting features from segmented objects. Third, we employ visual encoders to extract patch-level image features for constructing a patch-level memory bank for structural anomaly detection. These three complementary memory banks are used to retrieve and compare normal images that are most similar to the query image, compute anomaly scores at multiple levels, and fuse them into a final anomaly score. By unifying structural and logical anomaly detection through collaborative memory banks, TMUAD achieves state-of-the-art performance across seven publicly available datasets involving industrial and medical domains. The model and code are available at https://github.com/SIA-IDE/TMUAD.

Figures

Figures reproduced from arXiv: 2508.21795 by the authors.

Figure 1
Figure 1. Comparison between our TMUAD and existing frameworks. (a) The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The proposed Three-Memory framework for Unified Structural and logical Anomaly Detection (TMUAD). Our TMUAD incorporates three [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Proposed logic-aware text extractor for logical anomaly detection. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: Fusion of anomaly scores from three different memory banks. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The proposed pixel-level anomaly score calculation method for [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Representative examples from industrial [4]–[6], [51] and medical datasets [52]–[54] for both structural and logical anomalies with anomalous regions [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Visual comparison results on the MVTec LOCO [6] dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Visual comparison results on the MVTec AD [4] dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Visual comparison results on the VisA [5] dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Visual comparison results on the MPDD [51] dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 12
Figure 12. Figure 12: The relative contributions of three distinct memory banks to the [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: VLM’s advantage in providing fuzzy locations for object categories [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]
Figure 14
Figure 14. Figure 14: Logical anomaly detection by our text memory banks [PITH_FULL_IMAGE:figures/full_fig_p011_14.png]
Figure 17
Figure 17. Figure 17: Failure cases of our method on the MVTec LOCO [6] dataset. [PITH_FULL_IMAGE:figures/full_fig_p011_17.png]
Figure 16
Figure 16. Figure 16: Visual results for ablation studies of the patch-level image memory [PITH_FULL_IMAGE:figures/full_fig_p011_16.png]
Figure 18
Figure 18. Figure 18: Ablation study of the number of clusters in K-means of [PITH_FULL_IMAGE:figures/full_fig_p012_18.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

72 extracted references · 66 canonical work pages

  1. [1]

    Deep anomaly detection using geometric transformations,

    I. Golan and R. El-Yaniv, “Deep anomaly detection using geometric transformations,” in Proc. NeurIPS, vol. 31, 2018

    work page 2018

  2. [2]

    Towards total recall in industrial anomaly detection,

    K. Roth, L. Pemula, J. Zepeda, B. Sch ¨olkopf, T. Brox, and P. Gehler, “Towards total recall in industrial anomaly detection,” in Proc. CVPR, pp. 14318–14328, 2022

    work page 2022

  3. [3]

    Promptad: Learning prompts with only normal samples for few-shot anomaly detection,

    X. Li, Z. Zhang, X. Tan, C. Chen, Y . Qu, Y . Xie, and L. Ma, “Promptad: Learning prompts with only normal samples for few-shot anomaly detection,” in Proc. CVPR, pp. 16848–16858, 2024

    work page 2024

  4. [4]

    Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection,

    P. Bergmann, M. Fauser, D. Sattlegger, and C. Steger, “Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection,” in Proc. CVPR, pp. 9592–9600, 2019

    work page 2019

  5. [5]

    Spot-the- difference self-supervised pre-training for anomaly detection and seg- mentation,

    Y . Zou, J. Jeong, L. Pemula, D. Zhang, and O. Dabeer, “Spot-the- difference self-supervised pre-training for anomaly detection and seg- mentation,” in Proc. ECCV, pp. 392–408, 2022

    work page 2022

  6. [6]

    Be- yond dents and scratches: Logical constraints in unsupervised anomaly detection and localization,

    P. Bergmann, K. Batzner, M. Fauser, D. Sattlegger, and C. Steger, “Be- yond dents and scratches: Logical constraints in unsupervised anomaly detection and localization,” IJCV, vol. 130, no. 4, pp. 947–969, 2022

    work page 2022

  7. [7]

    Correcting deviations from normality: A reformulated diffusion model for multi- class unsupervised anomaly detection,

    F. Beizaee, G. A. Lodygensky, C. Desrosiers, and J. Dolz, “Correcting deviations from normality: A reformulated diffusion model for multi- class unsupervised anomaly detection,” in Proc. CVPR , pp. 19088– 19097, 2025. 13

    work page 2025

  8. [8]

    A cognitive memory-augmented network for visual anomaly detection.,

    T. Wang, X. Xu, F. Shen, and Y . Yang, “A cognitive memory-augmented network for visual anomaly detection.,” IEEE/CAA Journal of Automat- ica Sinica, vol. 8, no. 7, pp. 1296–1307, 2021

    work page 2021

Show all 72 references
  1. [9]

    Divide- and-assemble: Learning block-wise memory for unsupervised anomaly detection,

    J. Hou, Y . Zhang, Q. Zhong, D. Xie, S. Pu, and H. Zhou, “Divide- and-assemble: Learning block-wise memory for unsupervised anomaly detection,” in Proc. ICCV, pp. 8791–8800, 2021

    2021

  2. [10]

    Rethinking autoencoders for medical anomaly detection from a theoretical perspective,

    Y . Cai, H. Chen, and K.-T. Cheng, “Rethinking autoencoders for medical anomaly detection from a theoretical perspective,” in Proc. MICCAI , pp. 544–554, 2024

    2024

  3. [11]

    Realnet: A feature selection network with realistic synthetic anomaly for anomaly detection,

    X. Zhang, M. Xu, and X. Zhou, “Realnet: A feature selection network with realistic synthetic anomaly for anomaly detection,” in Proc. CVPR, pp. 16699–16708, 2024

    2024

  4. [12]

    A diffusion-based framework for multi-class anomaly detection,

    H. He, J. Zhang, H. Chen, X. Chen, Z. Li, X. Chen, Y . Wang, C. Wang, and L. Xie, “A diffusion-based framework for multi-class anomaly detection,” in Proc. AAAI, vol. 38, pp. 8472–8480, 2024

    2024

  5. [13]

    Residual denoising diffusion models,

    J. Liu, Q. Wang, H. Fan, Y . Wang, Y . Tang, and L. Qu, “Residual denoising diffusion models,” in Proc. CVPR, pp. 2773–2783, 2024

    2024

  6. [14]

    Glad: Towards better reconstruction with global and local adaptive diffusion models for unsupervised anomaly detection,

    H. Yao, M. Liu, Z. Yin, Z. Yan, X. Hong, and W. Zuo, “Glad: Towards better reconstruction with global and local adaptive diffusion models for unsupervised anomaly detection,” in Proc. ECCV, pp. 1–17, 2024

    2024

  7. [15]

    Anomaly detection via reverse distillation from one-class embedding,

    H. Deng and X. Li, “Anomaly detection via reverse distillation from one-class embedding,” in Proc. CVPR, pp. 9737–9746, 2022

    2022

  8. [16]

    Dual-modeling decouple distillation for unsupervised anomaly detection,

    X. Liu, J. Wang, B. Leng, and S. Zhang, “Dual-modeling decouple distillation for unsupervised anomaly detection,” in Proc. ACMMM , pp. 5035–5044, 2024

    2024

  9. [17]

    Feature-constrained and attention-conditioned distillation learning for visual anomaly detection,

    S. Zhang and J. Liu, “Feature-constrained and attention-conditioned distillation learning for visual anomaly detection,” in Proc. ICASSP , pp. 2945–2949, 2024

    2024

  10. [18]

    Aekd: Unsupervised auto- encoder knowledge distillation for industrial anomaly detection,

    Q. Wu, H. Li, C. Tian, L. Wen, and X. Li, “Aekd: Unsupervised auto- encoder knowledge distillation for industrial anomaly detection,”Journal of Manufacturing Systems , vol. 73, pp. 159–169, 2024

    2024

  11. [19]

    Pushing the limits of fewshot anomaly detection in industry vision: Graphcore,

    G. Xie, J. Wang, J. Liu, F. Zheng, and Y . Jin, “Pushing the limits of fewshot anomaly detection in industry vision: Graphcore,” in Proc. ICLR, 2023

    2023

  12. [20]

    Pni: industrial anomaly detection using position and neighborhood information,

    J. Bae, J.-H. Lee, and S. Kim, “Pni: industrial anomaly detection using position and neighborhood information,” in Proc. ICCV, pp. 6373–6383, 2023

    2023

  13. [21]

    Progressive bound- ary guided anomaly synthesis for industrial anomaly detection,

    Q. Chen, H. Luo, H. Gao, C. Lv, and Z. Zhang, “Progressive bound- ary guided anomaly synthesis for industrial anomaly detection,” IEEE TCSVT, vol. 35, no. 2, pp. 1193–1208, 2025

    2025

  14. [22]

    Inter-realization channels: Unsupervised anomaly detection beyond one-class classification,

    D. McIntosh and A. B. Albu, “Inter-realization channels: Unsupervised anomaly detection beyond one-class classification,” in Proc. ICCV , pp. 6262–6272, 2023

    2023

  15. [23]

    Reconpatch: Contrastive patch representation learning for industrial anomaly detection,

    J. Hyun, S. Kim, G. Jeon, S. H. Kim, K. Bae, and B. J. Kang, “Reconpatch: Contrastive patch representation learning for industrial anomaly detection,” in Proc. WACV, pp. 2052–2061, 2024

  16. [24]

    A reconstruction-based feature adaptation for anomaly detection with self-supervised multi-scale aggregation,

    Z. Zuo, Z. Wu, B. Chen, and X. Zhong, “A reconstruction-based feature adaptation for anomaly detection with self-supervised multi-scale aggregation,” in Proc. ICASSP, pp. 5840–5844, 2024

    2024

  17. [25]

    Towards training-free anomaly detection with vision and language foundation models,

    J. Zhang, G. Wang, Y . Jin, and D. Huang, “Towards training-free anomaly detection with vision and language foundation models,” in Proc. CVPR, pp. 15204–15213, 2025

    2025

  18. [26]

    Space: Spatial- aware consistency regularization for anomaly detection in industrial applications,

    D. Kim, H. Kim, D. Jeong, S. Suh, and H. Cho, “Space: Spatial- aware consistency regularization for anomaly detection in industrial applications,” in Proc. WACV, pp. 7184–7194, 2025

    2025

  19. [27]

    Contextual affinity distillation for image anomaly detection,

    J. Zhang, M. Suganuma, and T. Okatani, “Contextual affinity distillation for image anomaly detection,” in Proc. WACV, pp. 148–157, 2024

    2024

  20. [28]

    Efficientad: Accurate visual anomaly detection at millisecond-level latencies,

    K. Batzner, L. Heckler, and R. K ¨onig, “Efficientad: Accurate visual anomaly detection at millisecond-level latencies,” in Proc. WACV , pp. 127–137, 2024

    2024

  21. [29]

    Few shot part segmentation reveals compositional logic for industrial anomaly detection,

    S. Kim, S. An, P. Chikontwe, M. Kang, E. Adeli, K. M. Pohl, and S. H. Park, “Few shot part segmentation reveals compositional logic for industrial anomaly detection,” in Proc. AAAI, vol. 38, pp. 8591–8599, 2024

    2024

  22. [30]

    Univad: A training-free unified model for few-shot visual anomaly detection,

    Z. Gu, B. Zhu, G. Zhu, Y . Chen, M. Tang, and J. Wang, “Univad: A training-free unified model for few-shot visual anomaly detection,” in Proc. CVPR, pp. 15194–15203, 2025

    2025

  23. [31]

    Sam- lad: Segment anything model meets zero-shot logic anomaly detection,

    Y . Peng, X. Lin, N. Ma, J. Du, C. Liu, C. Liu, and Q. Chen, “Sam- lad: Segment anything model meets zero-shot logic anomaly detection,” Knowledge-Based Systems, vol. 314, p. 113176, 2025

    2025

  24. [32]

    Visual anomaly detection via partition memory bank module and error estimation,

    P. Xing and Z. Li, “Visual anomaly detection via partition memory bank module and error estimation,” IEEE TCSVT , vol. 33, no. 8, pp. 3596– 3607, 2023

    2023

  25. [33]

    Outlier-probability-based feature adaptation for robust unsupervised anomaly detection on contaminated training data,

    J. Zhou and Y . Wu, “Outlier-probability-based feature adaptation for robust unsupervised anomaly detection on contaminated training data,” IEEE TCSVT, vol. 34, no. 10, pp. 10023–10035, 2024

    2024

  26. [34]

    Gaussian mixture models,

    D. Reynolds, “Gaussian mixture models,” in Encyclopedia of biometrics, pp. 827–832, 2015

    2015

  27. [35]

    Anomalygpt: Detecting industrial anomalies using large vision-language models,

    Z. Gu, B. Zhu, G. Zhu, Y . Chen, M. Tang, and J. Wang, “Anomalygpt: Detecting industrial anomalies using large vision-language models,” in Proc. AAAI, vol. 38, pp. 1932–1940, 2024

    1932

  28. [36]

    Focusclip: Focusing on anomaly regions by visual-text discrepancies,

    Y . Zhao, J. Sun, L. Zhang, and H. Lu, “Focusclip: Focusing on anomaly regions by visual-text discrepancies,” IEEE TCSVT , vol. 35, no. 5, pp. 4883–4895, 2025

    2025

  29. [37]

    Crepe: Can vision-language foundation models reason compositionally?,

    Z. Ma, J. Hong, M. O. Gul, M. Gandhi, I. Gao, and R. Krishna, “Crepe: Can vision-language foundation models reason compositionally?,” in Proc. CVPR, pp. 10910–10921, 2023

    2023

  30. [38]

    Aa-clip: Enhancing zero-shot anomaly detection via anomaly-aware clip,

    W. Ma, X. Zhang, Q. Yao, F. Tang, C. Wu, Y . Li, R. Yan, Z. Jiang, and S. K. Zhou, “Aa-clip: Enhancing zero-shot anomaly detection via anomaly-aware clip,” in Proc. CVPR, pp. 4744–4754, 2025

    2025

  31. [39]

    Compositional chain- of-thought prompting for large multimodal models,

    C. Mitra, B. Huang, T. Darrell, and R. Herzig, “Compositional chain- of-thought prompting for large multimodal models,” in Proc. CVPR , pp. 14420–14431, 2024

    2024

  32. [40]

    Logicqa: Logical anomaly detection with vision language model generated questions,

    Y . Kwon, D. Moon, Y . Oh, and H. Yoon, “Logicqa: Logical anomaly detection with vision language model generated questions,” in Proc. ACL, 2025

    2025

  33. [41]

    Inves- tigating compositional challenges in vision-language models for visual grounding,

    Y . Zeng, Y . Huang, J. Zhang, Z. Jie, Z. Chai, and L. Wang, “Inves- tigating compositional challenges in vision-language models for visual grounding,” in Proc. CVPR, pp. 14141–14151, 2024

    2024

  34. [42]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo,et al., “Segment anything,” in Proc. ICCV, pp. 4015–4026, 2023

    2023

  35. [43]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al., “Learning transferable visual models from natural language supervision,” in Proc. ICML , pp. 8748–8763, 2021

    2021

  36. [44]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution,

    P. Wang, S. Bai, S. Tan, S. Wang, Z. Fan, J. Bai, K. Chen, X. Liu, J. Wang, W. Ge, et al., “Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution,” arXiv preprint arXiv:2409.12191, 2024

    2024 arXiv

  37. [45]

    Grounded sam: Assembling open-world models for diverse visual tasks,

    T. Ren, S. Liu, A. Zeng, J. Lin, K. Li, H. Cao, J. Chen, X. Huang, Y . Chen, F. Yan,et al., “Grounded sam: Assembling open-world models for diverse visual tasks,” arXiv preprint arXiv:2401.14159 , 2024

    2024 arXiv

  38. [46]

    Towards zero- shot anomaly detection and reasoning with multimodal large language models,

    J. Xu, S.-Y . Lo, B. Safaei, V . M. Patel, and I. Dwivedi, “Towards zero- shot anomaly detection and reasoning with multimodal large language models,” in Proc. CVPR, pp. 20370–20382, 2025

    2025

  39. [47]

    Gpt-4o system card,

    A. Hurst, A. Lerer, A. P. Goucher, A. Perelman, A. Ramesh, A. Clark, A. Ostrow, A. Welihinda, A. Hayes, A. Radford, et al., “Gpt-4o system card,” arXiv preprint arXiv:2410.21276 , 2024

    2024 arXiv

  40. [48]

    Mitigating hallucination in large multi-modal models via robust instruction tuning,

    F. Liu, K. Lin, L. Li, J. Wang, Y . Yacoob, and L. Wang, “Mitigating hallucination in large multi-modal models via robust instruction tuning,” 2024

    2024

  41. [49]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Nouby, et al. , “Dinov2: Learning robust visual features without supervision,” TMLR, 2023

    2023

  42. [50]

    From clip to dino: Visual encoders shout in multi-modal large language models,

    D. Jiang, Y . Liu, S. Liu, J. Zhao, H. Zhang, Z. Gao, X. Zhang, J. Li, and H. Xiong, “From clip to dino: Visual encoders shout in multi-modal large language models,” arXiv preprint arXiv:2310.08825 , 2023

    2023 arXiv

  43. [51]

    Deep learning-based defect detection of metal parts: evaluating current meth- ods in complex conditions,

    S. Jezek, M. Jonak, R. Burget, P. Dvorak, and M. Skotak, “Deep learning-based defect detection of metal parts: evaluating current meth- ods in complex conditions,” in Proc. ICUMT, pp. 66–71, 2021

    2021

  44. [52]

    The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification,

    U. Baid, S. Ghodasara, S. Mohan, M. Bilello, E. Calabrese, E. Colak, K. Farahani, J. Kalpathy-Cramer, F. C. Kitamura, S. Pati, et al. , “The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification,” arXiv preprint arXiv:2107.02314 , 2021

    2021 arXiv

  45. [53]

    Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge,

    B. Landman, Z. Xu, J. Igelsias, M. Styner, T. Langerak, and A. Klein, “Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge,” in Proc. MICCAI, vol. 5, p. 12, 2015

    2015

  46. [54]

    Automated segmentation of macular edema in oct using deep neural networks,

    J. Hu, Y . Chen, and Z. Yi, “Automated segmentation of macular edema in oct using deep neural networks,” Medical Image Analysis , vol. 55, pp. 216–227, 2019

    2019

  47. [55]

    Gestalt pattern matching

    D. E. M. John W. Ratcliff, “Gestalt pattern matching.” https://en. wikipedia.org/wiki/Gestalt pattern matching, 1983

    1983

  48. [56]

    Logical: Towards logical anomaly synthesis for unsupervised anomaly localization,

    Y . Zhao, “Logical: Towards logical anomaly synthesis for unsupervised anomaly localization,” in Proc. CVPR, pp. 4022–4031, 2024

    2024

  49. [57]

    Draem-a discriminatively trained reconstruction embedding for surface anomaly detection,

    V . Zavrtanik, M. Kristan, and D. Sko ˇcaj, “Draem-a discriminatively trained reconstruction embedding for surface anomaly detection,” in Proc. ICCV, pp. 8330–8339, 2021

    2021

  50. [58]

    Omnial: A unified cnn framework for unsupervised anomaly localization,

    Y . Zhao, “Omnial: A unified cnn framework for unsupervised anomaly localization,” in Proc. CVPR, pp. 3924–3933, 2023

    2023

  51. [59]

    Generalad: Anomaly detection across domains by attending to distorted features,

    L. P. Str ¨ater, M. Salehi, E. Gavves, C. G. Snoek, and Y . M. Asano, “Generalad: Anomaly detection across domains by attending to distorted features,” in Proc. ECCV, pp. 448–465, 2024. 14

    2024

  52. [60]

    Winclip: Zero-/few-shot anomaly classification and segmentation,

    J. Jeong, Y . Zou, T. Kim, D. Zhang, A. Ravichandran, and O. Dabeer, “Winclip: Zero-/few-shot anomaly classification and segmentation,” in Proc. CVPR, pp. 19606–19616, 2023

    2023

  53. [61]

    X. Chen, Y . Han, and J. Zhang, “April-gan: A zero-/few-shot anomaly classification and segmentation method for cvpr 2023 vand workshop challenge tracks 1&2: 1st place on zero-shot ad and 4th place on few- shot ad,” arXiv preprint arXiv:2305.17382 , 2023

    2023 arXiv

  54. [62]

    Anomalyclip: Object- agnostic prompt learning for zero-shot anomaly detection,

    Q. Zhou, G. Pang, Y . Tian, S. He, and J. Chen, “Anomalyclip: Object- agnostic prompt learning for zero-shot anomaly detection,” in Proc. ICLR, 2023

    2023

  55. [63]

    Adapting visual-language models for generalizable anomaly detection in medical images,

    C. Huang, A. Jiang, J. Feng, Y . Zhang, X. Wang, and Y . Wang, “Adapting visual-language models for generalizable anomaly detection in medical images,” in Proc. CVPR, pp. 11375–11385, 2024

    2024

  56. [64]

    Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly detection,

    Y . Cao, J. Zhang, L. Frittoli, Y . Cheng, W. Shen, and G. Boracchi, “Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly detection,” in Proc. ECCV, pp. 55–72, 2024

    2024

  57. [65]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection,

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, Q. Jiang, C. Li, J. Yang, H. Su, et al., “Grounding dino: Marrying dino with grounded pre-training for open-set object detection,” in Proc. ECCV, pp. 38–55, 2024

    2024

  58. [66]

    Simplenet: A simple network for image anomaly detection and localization,

    Z. Liu, Y . Zhou, Y . Xu, and Z. Wang, “Simplenet: A simple network for image anomaly detection and localization,” in Proc. CVPR, pp. 20402– 20411, 2023

    2023

  59. [67]

    Exploring the limits of transfer learning with a unified text-to-text transformer,

    C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y . Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,” JMLR, vol. 21, no. 140, pp. 1– 67, 2020

    2020

  60. [68]

    Detecting human- object contact in images,

    Y . Chen, S. K. Dwivedi, M. J. Black, and D. Tzionas, “Detecting human- object contact in images,” in Proc. CVPR, pp. 17100–17110, 2023

    2023

  61. [69]

    Interactvlm: 3d interaction reasoning from 2d foundational models,

    S. K. Dwivedi, D. Anti ´c, S. Tripathi, O. Taheri, C. Schmid, M. J. Black, and D. Tzionas, “Interactvlm: 3d interaction reasoning from 2d foundational models,” in Proc. CVPR, pp. 22605–22615, 2025

    2025

  62. [70]

    Product quantization for nearest neighbor search,

    H. Jegou, M. Douze, and C. Schmid, “Product quantization for nearest neighbor search,” IEEE TPAMI, vol. 33, no. 1, pp. 117–128, 2010. Jiawei Liu (Member, IEEE) received his B.E. degree from Northeast Agricultural University, Harbin, China, in 2018, and his Ph.D. degree in Patt...

    2010

  63. [2014]

    Her research interests include deep learning on image processing and medical image processing and applications

    She is currently a professor with the Institute of Shenyang Automation of the Chinese Academy of Sciences. Her research interests include deep learning on image processing and medical image processing and applications

  64. [2024]

    His research interests in- clude deep learning, illumination processing, image restoration, shadow removal, anomaly detection, and diffusion models

    He is currently a research assistant profes- sor at Shenyang Institute of Automation, Chinese Academy of Sciences. His research interests in- clude deep learning, illumination processing, image restoration, shadow removal, anomaly detection, and diffusion models. Jiahe Hou rec...

    2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.