Pith. sign in

REVIEW 5 major objections 5 minor 47 references

Bridge Feature Matching and Cross-Modal Alignment with Mutual-filtering for Zero-shot Anomaly Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a training-free CLIP pipeline, which uses the other images in the same test batch as filtered mutual references, achieves state-of-the-art zero-shot anomaly detection, reporting 93.3 AU-ROC for segmentation on…

desk verdict A genuinely new mutual-filtering mechanism for training-free zero-shot AD, with solid benchmark gains, but the paper overclaims practicality without testing high-anomaly batches or reporting batch size. read the letter →

arxiv 2507.11003 v1 pith:4EBPANPL submitted 2025-07-15 cs.CV

Yuhu Bai , Jiangning Zhang , Yunkang Cao , Guangyuan Lu , Qingdong He , Xiangtai Li , Guanzhong Tian This is my paper
classification cs.CV
keywords zero-shotanomalydetectionCLIPfeaturematchingcross-modalalignmentmutualfilteringbatchinferencesegmentationvision-languagemodels
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 aims to establish that zero-shot anomaly detection does not require fine-tuning, full-dataset priors, or hand-crafted prompts to be competitive: a frozen CLIP can detect and locate defects by using the other images in the same test batch as mutual references. Because those references are unlabeled and may themselves contain defects, the paper introduces a mutual-filtering loop that uses text-to-image alignment to mask out anomalous patches before feature matching, then feeds the matching scores back to refine the mask. On MVTec-AD this training-free recipe reports 93.3 AU-ROC and 49.1 F1-max for segmentation and 95.3 AU-ROC and 95.4 F1-max for classification, beating the previous training-free method MuSc-2 and the prompt-tuned AdaCLIP. If true, this matters because batch-based, training-free inference is closer to how industrial inspection actually runs, and it sidesteps the data-leakage risk of methods that adapt CLIP on auxiliary data.

What carries the argument

The central mechanism is the mutual-filtering loop between two CLIP-based heads. SeCLIP-AD computes per-patch anomaly probabilities by cosine similarity between CLIP visual features and averaged normal/abnormal text features; to restore local semantic correlation it replaces the final layer's q–k attention with the intermediate layer's attention weights and drops the residual connection and feed-forward network. FiCLIP-AD extracts patch tokens from layers 6, 12, 18, and 24, aggregates them over neighbourhoods $r = 1, 3, 5$, filters the reference features with the mask ($\hat{F}_v = F_v[\bar{M}]$), and scores each query patch by the minimum $\ell^2$ distance to the filtered reference set. The per-layer, per-scale scores are averaged into an intermediate mask that votes with the text-guided mask, and the loop repeats, so the same machinery both cleans the references and sharpens the anomaly map.

What would settle it

Batch the MVTec-AD test set into groups whose anomaly fraction is varied from 0% to 100% (for example, all-normal batches, natural batches, then batches made only of 'damaged' class images) and run the released FiSeCLIP code; if anomaly AU-ROC and F1-max fall monotonically and collapse near the all-defect end, the assumption that a batch always provides normal reference patches is falsified.

Watch

Extended reading notes

Core claim

FiSeCLIP's central claim is that zero-shot anomaly detection can be performed with a completely frozen CLIP by treating every test image in a batch as a reference for the others, provided the anomalous patches in those references are filtered out first. The filter is a text-guided mask: a patch is removed from the reference pool when its abnormal probability exceeds $\lambda$ times its normal probability, where those probabilities come from CLIP's similarity between visual patch features and averaged 'normal'/'abnormal' text features. After filtering, the score for each query patch is the minimum Euclidean distance to the remaining reference patches, and this score is then fed back to refine the mask in a mutual-filtering loop. The paper reports that this recipe reaches 93.3 AU-ROC and 49.1 F1-max for segmentation and 95.3 AU-ROC and 95.4 F1-max for classification on MVTec-AD, surpassing the previous training-free method MuSc-2 and the prompt-tuned AdaCLIP.

Load-bearing premise

The load-bearing premise is that, after text-guided filtering, the other images in a test batch still contain enough normal patches to serve as clean references; the paper never tests batches that are mostly or entirely defective, so a contaminated reference pool remains an unexamined failure mode.

Editorial extensions

If this is right

  • On MVTec-AD, FiSeCLIP reports 93.3 AU-ROC and 49.1 F1-max for segmentation and 95.3 AU-ROC and 95.4 F1-max for classification, exceeding the best training-free baseline MuSc-2 by 0.7 and 5.6 AU-ROC respectively.
  • On VisA, the method reports 97.4 AU-ROC segmentation, surpassing fine-tuned prompt methods such as AnomalyCLIP and the one-shot PatchCore baseline.
  • Because the model is training-free and the test distribution is not required in advance, it avoids the posterior optimization that MuSc needs and the data-leakage risk of prompt-tuning methods.
  • Batch-based inference means detection can run on the same groups of images that an industrial inspection line would process, with no need to hold the full test set in memory.
  • The mutual-filtering loop combines multi-layer and multi-scale features, so the method's gains come from the iterative refinement rather than any single similarity computation.

Reading between the lines

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

  • Editorial inference: on batches that are mostly or entirely defective, the mutual-filtering reference pool would collapse, so the method's reported gains likely depend on the low anomaly density of the benchmarks; this regime is not tested in the paper.
  • Editorial inference: because the text mask is prompt-driven, richer or class-specific prompts beyond plain 'normal' and 'abnormal' could plausibly improve filtering, and the paper only evaluates one fixed template.
  • Editorial inference: the batch requirement could be turned into a feature by deliberately injecting a known-good anchor image into every batch; the paper does not explore this, but the mechanism suggests it would stabilise the reference pool.
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

5 major / 5 minor

Summary. The paper proposes FiSeCLIP, a training-free zero-shot anomaly detection (ZSAD) method that uses a batch of test images as mutual references. It combines feature matching (FiCLIP-AD) with text-image alignment (SeCLIP-AD): text-guided masks filter out anomalous features from the reference images before computing patchwise minimum distances, and the resulting anomaly scores are iteratively fed back to refine the masks via mutual voting. The method also modifies the final CLIP attention layer by substituting intermediate-layer attention and removing the feed-forward network to improve local semantic coherence. Experiments on MVTec AD, VisA, and BTAD report state-of-the-art results, e.g., 93.3 AU-ROC and 49.1 F1-max for anomaly segmentation on MVTec AD, outperforming prior ZSAD methods including MuSc-2 and AdaCLIP.

Significance. If the claims are reproducible and the method generalizes beyond the specific evaluation protocol, FiSeCLIP is a valuable contribution to ZSAD: it is training-free, aligns with batch-based industrial deployment, and the mutual-filtering idea is a plausible way to suppress noisy references. The paper ships code, uses standard benchmarks, and includes ablations on the filtering mechanism, hyperparameters, and attention modifications. However, the reported superiority currently rests on an incompletely specified evaluation protocol: the batch size is never stated, hyperparameters are tuned on the test set, and the behavior under high anomaly prevalence is untested. These issues must be resolved before the SOTA claim can be taken at face value.

major comments (5)
  1. [Sec. 3.1, 4.1, Algorithm 1] The batch size B is never specified in the method or the implementation details. Algorithm 1 and Eq. (12) are written explicitly for two images (I_u and I_v), and the discussion of references uses v != u without explaining how multiple references are aggregated. If all experiments use B=2, this should be stated, and the comparison to MuSc-2 (which also uses two images) is then fair. If B>2 is used or intended, the mutual-filtering extension to multiple references is undefined, and the reported improvements over MuSc-2 would need to be disentangled from the added reference images. This is load-bearing for the central benchmark claim.
  2. [Sec. 4.3, Tables 3-5] The hyperparameters λ in Eq. (9) and μ in Eq. (13), as well as the attention layer index (Table 5), are selected by scanning values on the MVTec AD test set, with no validation split or fixed defaults used for the main results. This constitutes test-set adaptation, which is in tension with the 'zero-shot' and 'training-free' framing. The authors should report results for the default values without test-set tuning, or perform hyperparameter selection on a separate validation split (e.g., a subset of categories) and then report test performance.
  3. [Sec. 3.3.2, Eq. (11)] The anomaly score in Eq. (11) is the minimum distance over the filtered reference features. If the text-guided mask (Eq. 9) removes all reference patches, the minimum over an empty set is undefined, and the paper provides no fallback for this degenerate case. Because the entire contribution is the mutual-filtering mechanism, the method's behavior when reference batches are mostly or entirely anomalous should be tested. The benchmark datasets have low anomaly proportions (roughly 11-27%), so this regime is never exercised. Provide experiments varying the anomaly ratio of the reference batch (e.g., 0%, 50%, 100% anomalous references) and describe how empty references are handled.
  4. [Sec. 4.1, Evaluation] The paper does not report how test batches are constructed in the experiments (e.g., random sampling, fixed order, or category-wise grouping), nor does it report any variance across batch compositions. Since the method's output depends on which images are in the batch, the reported single-number results may be sensitive to the specific batch composition. Please provide error bars (mean and standard deviation) over multiple random batch compositions, or at minimum describe the batch construction in detail so the experiments can be reproduced.
  5. [Sec. 4.3, Hyperparameter ablation] The sentence 'The results indicate that the best performance is achieved when λ is and µ is 2' is incomplete: the optimal value for λ is missing, and Table 4 lists μ values from 0.47 to 0.57, not 2. This needs to be corrected to an explicit statement such as 'λ = 1.10 and μ = 0.57' if that is what the tables show.
minor comments (5)
  1. [Sec. 1] There is a typo in the introduction: 'vlaue' should be 'value'.
  2. [Sec. 4.1] The text says 'Average Prevision'; this should be 'Average Precision'.
  3. [Abstract and Sec. 1] The phrase 'SOTA AdaCLIP' is ambiguous because AdaCLIP is a fine-tuned method. Specify that it is the previous state-of-the-art among ZSAD methods, or say 'fine-tuned AdaCLIP'.
  4. [Figure 2] The figure caption annotates 'FiCLIP-AD (Sec. 3.2)' and 'SeCLIP-AD (Sec. 3.1)', but in the text SeCLIP-AD is Section 3.2 and FiCLIP-AD is Section 3.3. These references should be corrected.
  5. [Algorithm 1] The input 'initial anomaly score ˆa' is not defined in the text around Eqs. (12)-(13). Clarify whether this is the score from Eq. (7) or from a previous iteration of the loop.

Circularity Check

0 steps flagged · score 2.0 of 10

No material circularity: the SOTA benchmark claim is grounded in frozen CLIP and external MuSc/PatchCore feature matching; self-citations are baselines only, and the mutual-filtering loop is an iterative algorithm, not a derivation that assumes its conclusion.

full rationale

FiSeCLIP's central result is an empirical benchmark evaluation, not a derivation from assumptions. SeCLIP-AD and FiCLIP-AD use independent signals: patch-level text-image cosine similarity from frozen CLIP (Eqs. 6-7) and minimum Euclidean distance between visual patch tokens of batch images (Eqs. 8-11), following the external MuSc/PatchCore approach. The Eq. 9 mask is a thresholded text-image score; Eq. 13 refines it by voting with the feature-matching score. This is an alternating denoising loop, not an equation whose output is defined as its input. The reported MVTec/VisA/BTAD numbers are compared against external baselines (MuSc-2, WinCLIP, AdaCLIP, etc.), and the pretrained CLIP backbone is an external, frozen anchor. Several cited works (AdaCLIP, APRIL-GAN, CLIP-AD) are by the present authors, but they appear only as baselines or background and are never used to justify the method's correctness or to forbid alternatives. The hyperparameters lambda and mu are tuned on MVTec, and the batch-reference assumption is not stress-tested under high anomaly prevalence; these are generalization/correctness caveats, not circularity. No load-bearing step reduces to its own input by construction.

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

The central claim relies on two tuned hyperparameters (lambda, mu), one architecture choice (attention layer 12), and a domain assumption about batch composition. No new physical or model entities are introduced.

free parameters (3)
  • lambda = 1.10 (best on MVTec test set)
    Threshold in Eq.9 for constructing the initial anomaly mask from text-image probabilities; tuned on the MVTec-AD test set in Table 3.
  • mu = 0.57 (default, best on MVTec test set)
    Threshold in Eq.13 for binary mask from average feature-match scores; tuned on MVTec-AD test set in Table 4.
  • attention_layer_index = 12
    Layer whose attention weights replace the final layer's attention in SeCLIP-AD; selected because it yields the best local-semantic restoration on MVTec-AD in Table 5.
assumptions (3)
  • domain assumption Test images within a batch are mostly normal and can serve as mutual references.
    Inherited from MuSc's observation that >95% of test pixels are normal; not explicitly validated for small batch sizes in this paper.
  • domain assumption Intermediate-layer attention in CLIP preserves local semantics better than the final layer for dense prediction.
    Empirically supported by Fig.3 and Table 5 on MVTec-AD, but not theoretically justified and not cross-validated on other datasets.
  • domain assumption Text prompts such as 'normal' and 'abnormal' can discriminate anomaly in CLIP's feature space.
    Standard assumption in CLIP-based ZSAD, cited from WinCLIP; the paper relies on it to generate the initial mask.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridge Feature Matching and Cross-Modal Alignment with Mutual-filtering for Zero-shot Anomaly Detection." pith.science (2026). https://pith.science/paper/4EBPANPL

@misc{pith2026250711003,
  author       = {Pith},
  title        = {Pith review of: Bridge Feature Matching and Cross-Modal Alignment with Mutual-filtering for Zero-shot Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4EBPANPL}},
  note         = {Machine review of arXiv:2507.11003}
}
abstract

With the advent of vision-language models (e.g., CLIP) in zero- and few-shot settings, CLIP has been widely applied to zero-shot anomaly detection (ZSAD) in recent research, where the rare classes are essential and expected in many applications. This study introduces \textbf{FiSeCLIP} for ZSAD with training-free \textbf{CLIP}, combining the feature matching with the cross-modal alignment. Testing with the entire dataset is impractical, while batch-based testing better aligns with real industrial needs, and images within a batch can serve as mutual reference points. Accordingly, FiSeCLIP utilizes other images in the same batch as reference information for the current image. However, the lack of labels for these references can introduce ambiguity, we apply text information to \textbf{fi}lter out noisy features. In addition, we further explore CLIP's inherent potential to restore its local \textbf{se}mantic correlation, adapting it for fine-grained anomaly detection tasks to enable a more accurate filtering process. Our approach exhibits superior performance for both anomaly classification and segmentation on anomaly detection benchmarks, building a stronger baseline for the direction, e.g., on MVTec-AD, FiSeCLIP outperforms the SOTA AdaCLIP by +4.6\%$\uparrow$/+5.7\%$\uparrow$ in segmentation metrics AU-ROC/$F_1$-max.

Figures

Figures reproduced from arXiv: 2507.11003 by the authors.

Figure 1
Figure 1. Compared with (a) mainstream ZSAD methods [ [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of our FiSeCLIP. It consists of two parts: SeCLIP-AD (Sec. 3.2) and FiCLIP-AD (Sec. 3.3). 1) For two images in the same batch, SeCLIP-AD calculates the similarity between text and visual features to generate the corresponding anomaly score and mask, which are then used by the FiCLIP-AD module. The Proxy Layer is used to restore semantic correlation among patches, with attention weights derived from the inte… view at source ↗
Figure 3
Figure 3. Attention maps from different layers of CLIP, with red dots indicating the selected patch positions in the image, which shows 12-th attention map has the best local semantics. and [c] indicates the image’s category. Two types of text prompts, normal and abnormal, are generated. We utilize CLIP’s text encoder to extract text features, and the final text features Ft ∈ R 2×C can be obtained by averaging the features of… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: An example of mutual-filtering. Iu denotes the inference image, Iv represents the reference image without filtering, and I ′ v represents the reference image with filtering. M. Specifically, the four stages patch tokens are utilized to generate the average anomaly scor…
Figure 5
Figure 5. Figure 5: Visualization results of anomaly localization for each class on MVTec AD (left) dataset and VisA dataset (right) [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Histogram of anomaly classification scores on the [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 31 canonical work pages

  1. [2]

    Pni: Indus- trial anomaly detection using position and neighborhood in- formation

    Jaehyeok Bae, Jae-Han Lee, and Seyun Kim. Pni: Indus- trial anomaly detection using position and neighborhood in- formation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023. 2

    work page 2023

  2. [3]

    Dual-path frequency discriminators for few-shot anomaly detection

    Yuhu Bai, Jiangning Zhang, Zhaofeng Chen, Yuhang Dong, Yunkang Cao, and Guanzhong Tian. Dual-path frequency discriminators for few-shot anomaly detection. Knowledge- Based Systems, 2024. 2

    work page 2024

  3. [4]

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

    Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit , pages 9592–9600,

    work page

  4. [5]

    The mvtec anomaly detection dataset: a comprehensive real-world dataset for unsupervised anomaly detection

    Paul Bergmann, Kilian Batzner, Michael Fauser, David Sat- tlegger, and Carsten Steger. The mvtec anomaly detection dataset: a comprehensive real-world dataset for unsupervised anomaly detection. International Journal of Computer Vi- sion, 2021. 6

    work page 2021

  5. [6]

    A sur- vey on visual anomaly detection: Challenge, approach, and prospect

    Yunkang Cao, Xiaohao Xu, Jiangning Zhang, Yuqi Cheng, Xiaonan Huang, Guansong Pang, and Weiming Shen. A sur- vey on visual anomaly detection: Challenge, approach, and prospect. arXiv:2401.16402, 2024. 1

    arXiv 2024

  6. [7]

    Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly de- tection

    Yunkang Cao, Jiangning Zhang, Luca Frittoli, Yuqi Cheng, Weiming Shen, and Giacomo Boracchi. Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly de- tection. In European Conference on Computer Vision, 2025. 2, 3, 6

    work page 2025

  7. [8]

    A unified anomaly synthesis strategy with gradi- ent ascent for industrial anomaly detection and localization

    Qiyu Chen, Huiyuan Luo, Chengkan Lv, and Zhengtao Zhang. A unified anomaly synthesis strategy with gradi- ent ascent for industrial anomaly detection and localization. arXiv preprint arXiv:2407.09359, 2024. 2

    arXiv 2024

  8. [9]

    Xuhai Chen, Yue Han, and Jiangning 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. 1, 3, 6, 7

    arXiv 2023

Show all 47 references
  1. [10]

    Clip-ad: A language-guided staged dual-path model for zero-shot anomaly detection

    Xuhai Chen, Jiangning Zhang, Guanzhong Tian, Haoyang He, Wuhao Zhang, Yabiao Wang, Chengjie Wang, and Yong Liu. Clip-ad: A language-guided staged dual-path model for zero-shot anomaly detection. arXiv preprint arXiv:2311.00453, 2023. 1, 3

    2023 arXiv

  2. [11]

    Padim: a patch distribution modeling framework for anomaly detection and localization

    Thomas Defard, Aleksandr Setkov, Angelique Loesch, and Romaric Audigier. Padim: a patch distribution modeling framework for anomaly detection and localization. In Proc. Int. Conf. Pattern. Recognit, 2021. 2

    2021

  3. [12]

    Anomaly detection via reverse distillation from one-class embedding

    Hanqiu Deng and Xingyu Li. Anomaly detection via reverse distillation from one-class embedding. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit, 2022. 1

    2022

  4. [13]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 2

    2010 arXiv

  5. [14]

    Filo: Zero-shot anomaly detection by fine-grained description and high-quality local- ization

    Zhaopeng Gu, Bingke Zhu, Guibo Zhu, Yingying Chen, Hao Li, Ming Tang, and Jinqiao Wang. Filo: Zero-shot anomaly detection by fine-grained description and high-quality local- ization. arXiv preprint arXiv:2404.13671, 2024. 1, 2, 3

    2024

  6. [15]

    Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows

    Denis Gudovskiy, Shun Ishizaka, and Kazuki Kozuka. Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows. In Proc. IEEE Winter Conf. Appl. Comput. Vis., 2022. 2

    2022

  7. [16]

    Mambaad: Exploring state space models for multi-class unsupervised anomaly detec- tion

    Haoyang He, Yuhu Bai, Jiangning Zhang, Qingdong He, Hongxu Chen, Zhenye Gan, Chengjie Wang, Xiangtai Li, Guanzhong Tian, and Lei Xie. Mambaad: Exploring state space models for multi-class unsupervised anomaly detec- tion. arXiv preprint arXiv:2404.06564, 2024. 2, 6

    2024 arXiv

  8. [17]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 2

    2016

  9. [18]

    Anomalyd- iffusion: Few-shot anomaly image generation with diffusion model

    Teng Hu, Jiangning Zhang, Ran Yi, Yuzhen Du, Xu Chen, Liang Liu, Yabiao Wang, and Chengjie Wang. Anomalyd- iffusion: Few-shot anomaly image generation with diffusion model. In Proc. AAAI Conf. Artif. Intell., 2024. 2

    2024

  10. [19]

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

    Jongheon Jeong, Yang Zou, Taewan Kim, Dongqing Zhang, Avinash Ravichandran, and Onkar Dabeer. Winclip: Zero- /few-shot anomaly classification and segmentation. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit , pages 19606– 19616, 2023. 1, 3, 6, 7

    2023

  11. [20]

    Clearclip: Decom- posing clip representations for dense vision-language infer- ence

    Mengcheng Lan, Chaofeng Chen, Yiping Ke, Xinjiang Wang, Litong Feng, and Wayne Zhang. Clearclip: Decom- posing clip representations for dense vision-language infer- ence. arXiv preprint arXiv:2407.12442, 2024. 2, 3

    2024 arXiv

  12. [21]

    Proxyclip: Proxy attention improves clip for open-vocabulary segmentation

    Mengcheng Lan, Chaofeng Chen, Yiping Ke, Xinjiang Wang, Litong Feng, and Wayne Zhang. Proxyclip: Proxy attention improves clip for open-vocabulary segmentation. arXiv preprint arXiv:2408.04883, 2024. 2

    2024 arXiv

  13. [22]

    Pyramid- flow: High-resolution defect contrastive localization using pyramid normalizing flow

    Jiarui Lei, Xiaobo Hu, Yue Wang, and Dong Liu. Pyramid- flow: High-resolution defect contrastive localization using pyramid normalizing flow. InProc. IEEE Conf. Comput. Vis. Pattern Recognit, pages 14143–14152, 2023. 1

    2023

  14. [23]

    Zero-shot anomaly detection via batch normalization

    Aodong Li, Chen Qiu, Marius Kloft, Padhraic Smyth, Maja Rudolph, and Stephan Mandt. Zero-shot anomaly detection via batch normalization. In NeurIPS, 2023. 6

    2023

  15. [24]

    Zero-shot anomaly detection via batch normalization

    Aodong Li, Chen Qiu, Marius Kloft, Padhraic Smyth, Maja Rudolph, and Stephan Mandt. Zero-shot anomaly detection via batch normalization. Advances in Neural Information Processing Systems, 2023. 1

    2023

  16. [25]

    Cutpaste: Self-supervised learning for anomaly de- tection and localization

    Chun-Liang Li, Kihyuk Sohn, Jinsung Yoon, and Tomas Pfister. Cutpaste: Self-supervised learning for anomaly de- tection and localization. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit, 2021. 2

    2021

  17. [26]

    Musc: Zero-shot industrial anomaly classification and segmentation with mutual scoring of the unlabeled images

    Xurui Li, Ziming Huang, Feng Xue, and Yu Zhou. Musc: Zero-shot industrial anomaly classification and segmentation with mutual scoring of the unlabeled images. arXiv preprint arXiv:2401.16753, 2024. 1, 2, 3, 5, 6, 7

    2024 arXiv

  18. [27]

    Clip surgery for better explainability with enhancement in open- vocabulary tasks

    Yi Li, Hualiang Wang, Yiqun Duan, and Xiaomeng Li. Clip surgery for better explainability with enhancement in open- vocabulary tasks. arXiv preprint arXiv:2304.05653 , 2023. 2

    2023 arXiv

  19. [28]

    Omni-frequency channel- 9 selection representations for unsupervised anomaly detec- tion

    Yufei Liang, Jiangning Zhang, Shiwei Zhao, Runze Wu, Yong Liu, and Shuwen Pan. Omni-frequency channel- 9 selection representations for unsupervised anomaly detec- tion. IEEE Trans. Image Process., 2023. 2

    2023

  20. [29]

    Deep indus- trial image anomaly detection: A survey

    Jiaqi Liu, Guoyang Xie, Jinbao Wang, Shangnian Li, Chengjie Wang, Feng Zheng, and Yaochu Jin. Deep indus- trial image anomaly detection: A survey. Machine Intelli- gence Research, 2024. 1

    2024

  21. [30]

    Deep indus- trial image anomaly detection: A survey

    Jiaqi Liu, Guoyang Xie, Jinbao Wang, Shangnian Li, Chengjie Wang, Feng Zheng, and Yaochu Jin. Deep indus- trial image anomaly detection: A survey. Mach. Intell. Res., 21:104–135, 2024. 1

    2024

  22. [31]

    Reb: Re- ducing biases in representation for industrial anomaly detec- tion

    Shuai Lyu, Dongmei Mo, and Wai keung Wong. Reb: Re- ducing biases in representation for industrial anomaly detec- tion. Knowledge-Based Syst., 290:111563, 2024. 2

    2024

  23. [32]

    Vt-adl: A vision trans- former network for image anomaly detection and localiza- tion

    Pankaj Mishra, Riccardo Verk, Daniele Fornasier, Claudio Piciarelli, and Gian Luca Foresti. Vt-adl: A vision trans- former network for image anomaly detection and localiza- tion. In 2021 IEEE 30th International Symposium on Indus- trial Electronics (ISIE), 2021. 6

    2021

  24. [33]

    Vcp-clip: A visual context prompting model for zero-shot anomaly segmenta- tion

    Zhen Qu, Xian Tao, Mukesh Prasad, Fei Shen, Zhengtao Zhang, Xinyi Gong, and Guiguang Ding. Vcp-clip: A visual context prompting model for zero-shot anomaly segmenta- tion. arXiv preprint arXiv:2407.12276, 2024. 3

    2024 arXiv

  25. [34]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  26. [35]

    Towards total recall in industrial anomaly detection

    Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Sch¨olkopf, Thomas Brox, and Peter Gehler. Towards total recall in industrial anomaly detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit, 2022. 1, 2, 5, 6

    2022

  27. [36]

    Opti- mizing patchcore for few/many-shot anomaly detection

    Jo ˜ao Santos, Triet Tran, and Oliver Rippel. Opti- mizing patchcore for few/many-shot anomaly detection. arXiv:2307.10792, 2023. 5

    2023 arXiv

  28. [37]

    Ex- plore the potential of clip for training-free open vocabulary semantic segmentation

    Tong Shao, Zhuotao Tian, Hang Zhao, and Jingyong Su. Ex- plore the potential of clip for training-free open vocabulary semantic segmentation. In European Conference on Com- puter Vision, pages 139–156. Springer, 2025. 2

    2025

  29. [38]

    Attention guided anomaly localization in images

    Shashanka Venkataramanan, Kuan-Chuan Peng, Ra- jat Vikram Singh, and Abhijit Mahalanobis. Attention guided anomaly localization in images. In European Conference on Computer Vision, 2020. 1, 2

    2020

  30. [39]

    Sclip: Rethinking self-attention for dense vision-language inference

    Feng Wang, Jieru Mei, and Alan Yuille. Sclip: Rethinking self-attention for dense vision-language inference. In Euro- pean Conference on Computer Vision, 2025. 2

    2025

  31. [40]

    Anoddpm: Anomaly detection with de- noising diffusion probabilistic models using simplex noise

    Julian Wyatt, Adam Leach, Sebastian M Schmon, and Chris G Willcocks. Anoddpm: Anomaly detection with de- noising diffusion probabilistic models using simplex noise. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit , pages 650–656, 2022. 2

    2022

  32. [41]

    A unified model for multi-class anomaly detection

    Zhiyuan You, Lei Cui, Yujun Shen, Kai Yang, Xin Lu, Yu Zheng, and Xinyi Le. A unified model for multi-class anomaly detection. In Proc. Adv. Neural Inf. Process. Syst., pages 4571–4584, 2022. 1

    2022

  33. [42]

    Draem- a discriminatively trained reconstruction embedding for sur- face anomaly detection

    Vitjan Zavrtanik, Matej Kristan, and Danijel Skoˇcaj. Draem- a discriminatively trained reconstruction embedding for sur- face anomaly detection. In Proc. IEEE Int. Conf. Comput. Vis., 2021. 2, 6

    2021

  34. [43]

    Exploring plain vit reconstruction for multi- class unsupervised anomaly detection

    Jiangning Zhang, Xuhai Chen, Yabiao Wang, Chengjie Wang, Yong Liu, Xiangtai Li, Ming-Hsuan Yang, and Dacheng Tao. Exploring plain vit reconstruction for multi- class unsupervised anomaly detection. arXiv:2312.07495,

    arXiv

  35. [44]

    A comprehensive library for bench- marking multi-class visual anomaly detection.arXiv preprint arXiv:2406.03262, 2024

    Jiangning Zhang, Haoyang He, Zhenye Gan, Qingdong He, Yuxuan Cai, Zhucun Xue, Yabiao Wang, Chengjie Wang, Lei Xie, and Yong Liu. A comprehensive library for bench- marking multi-class visual anomaly detection.arXiv preprint arXiv:2406.03262, 2024. 1

    2024 arXiv

  36. [45]

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

    Ximiao Zhang, Min Xu, and Xiuzhuang Zhou. Realnet: A feature selection network with realistic synthetic anomaly for anomaly detection. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2024. 2

    2024

  37. [46]

    Dual-image enhanced clip for zero-shot anomaly detection

    Zhaoxiang Zhang, Hanqiu Deng, Jinan Bao, and Xingyu Li. Dual-image enhanced clip for zero-shot anomaly detection. arXiv preprint arXiv:2405.04782, 2024. 3, 6

    2024 arXiv

  38. [47]

    Anomalyclip: Object-agnostic prompt learn- ing for zero-shot anomaly detection

    Qihang Zhou, Guansong Pang, Yu Tian, Shibo He, and Jiming Chen. Anomalyclip: Object-agnostic prompt learn- ing for zero-shot anomaly detection. arXiv preprint arXiv:2310.18961, 2023. 1, 2, 3, 6, 7

    2023

  39. [48]

    Spot-the-difference self-supervised pre- training for anomaly detection and segmentation

    Yang Zou, Jongheon Jeong, Latha Pemula, Dongqing Zhang, and Onkar Dabeer. Spot-the-difference self-supervised pre- training for anomaly detection and segmentation. In Proc. Eur. Conf. Comput. Vis., pages 392–408. Springer, 2022. 2, 6, 7 10

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.