Pith. sign in

REVIEW 4 major objections 6 minor 52 references

CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A text 'normal' anchor lifts multi-class defect detection to 98.6

desk verdict A well-ablated integration of CLIP prompts and gated MoE into reverse distillation that delivers consistent gains on two AD benchmarks; the mechanism is plausible but under-tested. read the letter →

arxiv 2501.00346 v1 pith:NOS62LMD submitted 2024-12-31 cs.CV cs.AIcs.LG

Xiaolei Wang , Xiaoyang Wang , Huihui Bai , Eng Gee Lim , Jimin Xiao This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords unsupervisedanomalydetectionmulti-classreversedistillationcross-modalnormalityconstraintCLIPpromptlearningmixture-of-expertsindustrialdefectMVTecAD
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

Unsupervised anomaly detection via reverse distillation trains a decoder to reproduce normal image features, then flags test regions where the reconstruction fails. This paper argues that in multi-class training the decoder 'over-generalizes': because it sees far more varied normal patches, it also reconstructs abnormal patches too well, shrinking the very difference that reveals defects. The proposed remedy, the Cross-modal Normality Constraint (CNC), steers decoded features toward a class-agnostic textual representation of normality learned by CLIP, a vision-language model, and a gated mixture-of-experts module lets different patch patterns be processed by different experts. On the MVTec AD and VisA benchmarks, the combined framework reports image-level AUROC of 98.6 and 93.2 respectively, surpassing the single-modal multi-class distillation baselines RD4AD, UniAD, and DiAD. The result matters because it suggests one model can serve many product classes without the usual per-class accuracy penalty.

What carries the argument

The load-bearing mechanism is the cross-modal normality constraint built on CLIP. The paper uses class-agnostic learnable prompts $p_n$ ('normal object') and $p_a$ ('damaged object'), encoded by the frozen CLIP text encoder into text features $g_n$ and $g_a$. Two alignment losses, $\mathcal{L}^1_c$ and $\mathcal{L}^2_c$, push the global features of encoded and decoded visual patches toward $g_n$ rather than $g_a$; a dynamic anchor in text space therefore pulls the decoder back to a shared 'normal' direction. The feature-level normality promotion computes a control coefficient $\Psi(\alpha_i,\beta_i)=\tfrac12(1+\tanh(\alpha_i-\beta_i))$, where $\alpha_i$ and $\beta_i$ are patch-level similarities to $g_n$ and $g_a$, and adds $\lambda\Psi$ to encoded and decoded features before the distillation loss. The gated mixture-of-experts module takes a multi-layer fusion of encoded features and routes each patch to the top two of five expert MLPs, with an importance loss to keep routing balanced; the total loss is $\mathcal{L}_{\text{total}}=\mathcal{L}_{\text{distill}}+\mathcal{L}_{\text{constraint}}+\mathcal{L}_{\text{moe}}$.

What would settle it

For each MVTec AD class, compute the CLIP cosine similarity between the learned 'normal' prompt embedding and the encoded features of normal versus anomalous patches; if there is a class where anomalous patches are closer to the 'normal' text than normal patches are, the cross-modal normality assumption fails for that class and the constraint would be expected to hurt rather than help.

Watch

Extended reading notes

Core claim

The paper's central claim is that the 'over-generalization' (OG) of the decoder is the main obstacle in unsupervised multi-class distillation, and that it can be suppressed by a cross-modal normality constraint rather than by more visual data or per-class models. The authors show that normal and abnormal patches respond differently to the text descriptions 'a photo of a normal object' and 'a photo of a damaged object', and they exploit this asymmetry: learnable, class-agnostic prompts are passed through the CLIP text encoder to produce normality anchors, decoded features are pulled toward the normal anchor via a cross-modal constraint loss, and a feature-level normality promotion term reweights encoded and decoded patch features by a tanh-shaped control coefficient. A gated mixture-of-experts module, with five experts and top-two routing, reduces mutual interference among patch patterns from different classes. On MVTec AD the method reports 98.6 image-level AUROC, 98.0 pixel-level AUROC, and a mean mTotal of 89.0, and on VisA it reports 93.2 image-level AUROC and 98.5 pixel-level AUROC, improving over RD4AD, UniAD, and DiAD in the multi-class setting.

Load-bearing premise

The method assumes that the text phrases 'a photo of a normal object' and 'a photo of a damaged object' carry a reliable visual signal for every industrial class, so that steering decoded features toward the 'normal' text embedding improves detection instead of distorting it.

Editorial extensions

If this is right

  • One model trained jointly on many classes can match or beat single-class-trained specialists, cutting storage and training time for industrial inspection lines.
  • The normality anchor is class-agnostic, so a new product class can be added without designing new per-class text prompts.
  • The decoder's generalization to abnormal patches is not inevitable; it can be countered by an external semantic constraint rather than only by more normal training data.
  • Routing patches to expert subnetworks is a transferable way to reduce inter-class interference in other multi-domain feature-distillation tasks.
  • The reported per-class stability on MVTec AD means the method avoids the catastrophic per-class failures seen in baselines.

Reading between the lines

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

  • Per-class analysis of CLIP text-patch similarity would show whether the constraint helps most where 'normal' and 'damaged' prompts are well separated; on classes like Grid or Screw, the reported gain may come mostly from the MoE or from feature perturbation rather than from the text anchor.
  • By extension, the tanh control coefficient is a soft gating function, so replacing it with a learned scalar or a temperature-scaled sigmoid would test whether the improvement comes from the specific cross-modal form or simply from perturbing features.
  • Because the prompts are class-agnostic, the same framework could be applied zero-shot or few-shot by initializing prompts from a few normal images or skipping prompt learning entirely, then measuring detection on unseen classes.
  • The success of routing patch embeddings to different experts suggests that patch-pattern specialization could transfer to other multi-distribution settings, such as multi-domain segmentation or generalized out-of-distribution detection.
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 / 6 minor

Summary. The paper proposes CNC, an unsupervised multi-class anomaly detection method built on reverse distillation with a CLIP visual encoder. The method introduces class-agnostic learnable prompts that define normal and abnormal text features, a cross-modal normality constraint that aligns decoded features toward the normal text representation, a feature-level normality promotion mechanism that uses a text-derived control coefficient on encoded and decoded features, and a gated mixture-of-experts module to handle diverse patch patterns. The model is trained on normal samples only and detects anomalies by comparing encoded and decoded features. The authors report competitive results on MVTec AD (I-AUROC 98.6, mTotal 89.0) and VisA (I-AUROC 93.2, mTotal 82.7), with ablations supporting the individual components.

Significance. If the results hold, the paper addresses a real problem in multi-class unsupervised anomaly detection: the decoder over-generalizing to abnormal patches when trained across many classes. A strength is that the method is evaluated on two standard datasets with five metrics, compared against strong baselines (RD4AD, UniAD, DiAD), and accompanied by code. The ablations in Table 2 and the MoE hyperparameter analysis in Table 4 are useful. The central caveat is that the cross-modal text signal is the load-bearing premise, yet the paper does not validate per-class text alignment or compare against a random-text control; the mechanism of 'over-generalization' is also asserted rather than directly measured. The shape mismatch in Eq. (5) is a concrete reproducibility issue. With those concerns addressed, the contribution would be a solid step for multi-class unsupervised AD.

major comments (4)
  1. [Feature Distillation with Normality Promotion, Eqs. (5)-(8)] In Eq. (5), f_i is in R^{H x W x C} while Psi(alpha_i, beta_i) is in R^{H x W} according to the text below Eq. (6) and Eq. (7); the paper says the operator is element-wise addition, which is undefined for these shapes. Since f_i^* and bf_i^* are used in the distillation loss Eq. (9), the FNP mechanism as written is not reproducible. Please specify how the spatial control map is broadcast to the channel dimension or otherwise transformed before addition.
  2. [Learning Cross-modal Normality / Feature Distillation with Normality Promotion, Fig. 1(A)] The method's operating premise is that class-agnostic 'a photo of a normal object' and 'a photo of a damaged object' text features provide a per-patch normal-versus-abnormal signal that transfers across all 15 MVTec AD and 12 VisA classes. The paper supports this only with illustrative Fig. 1(A) and aggregate means; Eqs. (4)-(10) apply the text-based coefficient to every patch, so a class with weak cross-modal alignment could be hurt rather than helped. Please provide per-class text-alignment statistics (e.g., the distribution of alpha_i - beta_i on normal versus abnormal patches) and a control experiment that replaces the learned text features with fixed random vectors or permutations; without this, the reported mean gains do not distinguish a genuine cross-modal effect from a generic feature transformation that happens to help on the chosen benchmarks.
  3. [Introduction and Methodology (over-generalization claim)] The 'over-generalization' (OG) mechanism is asserted in the Introduction and repeated throughout the Methodology, but it is not directly measured. The paper does not quantify, for example, the reconstruction error of the vanilla decoder on held-out abnormal features versus normal features, nor does it show that the proposed constraint specifically reduces decoder success on abnormal patches rather than simply changing feature geometry. An ablation measuring patch-level reconstruction errors on normal and abnormal data with and without CNC would make the central claim testable; alternatively, the conclusions should be softened to describe the observed performance improvement without committing to the OG mechanism.
  4. [Experimental Setup / Implementation Details, Table 4] Several core hyperparameters (tau=0.001, gamma=0.1, theta=5, prompt length=12, and especially the MoE choices T=5 and K=2 from Table 4) appear to be selected using full test-set results, and no validation split or selection protocol is described. Since Table 4 shows that MoE performance varies by about 1.2 I-AUROC across configurations, reporting the best test configuration as the final result can produce an optimistic comparison with baselines. Please clarify how hyperparameters were chosen or report sensitivity on a held-out validation split.
minor comments (6)
  1. [Feature Distillation with Normality Promotion, Eq. (9)] The phrase 'fatten function' should be 'flatten function'.
  2. [Experimental Setup] In the experiments text, 'P-AURPOC' appears to be a typo for pixel-level AUROC; please correct it.
  3. [Learning Cross-modal Normality, Eqs. (4) and (10)] The global features e_i and be_i are used in the losses but never defined; please state whether they are CLS tokens, average-pooled features, or another pooling of f_i and bf_i.
  4. [Inference, Eq. (17)] In Eq. (17), sigma_i is called an upsampling factor but is used as an upsampling operation; please specify the interpolation method and the factor used to match the input resolution.
  5. [Gated Mixture-of-Experts, Eq. (15)] The 'universal importance loss' is cited to Bengio et al. 2015, but the form in Eq. (15) closely resembles the importance loss from Shazeer et al. 2017; please verify the citation.
  6. [Inference] The inference paragraph says the method applies the encoder, learned prompts, MLF, MoE module, and decoder, but it omits the CLIP text encoder G, which is needed to produce g_i^n and g_i^a for the feature-level normality promotion; please clarify the full inference pathway.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is trained on normal training data, evaluated on held-out test sets against external baselines, and no reported metric is defined in terms of the learned prompts or fitted parameters.

full rationale

The paper's central claim is an empirical benchmark result on MVTec AD and VisA, obtained by training the proposed cross-modal distillation framework on anomaly-free training images and evaluating on held-out test images with standard metrics (I-AUROC, P-AUROC, AUPRO, I-mAP, P-mAP). The training objective L_total = L_distill + L_constraint + L_moe is optimized on normal samples only; the anomaly score at inference, S = sum_i sigma_i (1 - d(f*_i, bf*_i)), is computed from encoded and decoded features without using test labels. The learnable text prompts are optimized via Eq. (4) on the training set and then used as anchors in Eqs. (5)-(11); they are not fitted to the target AUROC or to the test set. The 'over-generalization' hypothesis is an explanatory narrative rather than a mathematical claim derived from the model equations, so it cannot be circular by construction. Baseline numbers are sourced from external published work (RD4AD, UniAD, DiAD), not from the authors' own prior results, and no uniqueness theorem or self-citation is invoked to force the method choice. Hyperparameter choices such as T, K, and resolution are tuned on the benchmark, which is a tuning loop rather than a derivational circularity. No equation in the paper reduces a reported prediction to an input definition, and no fitted parameter is renamed as a prediction. The paper is therefore self-contained with respect to its empirical evaluation and exhibits no significant circularity.

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

The central claim rests on 5 fitted prompt/hyperparameter settings and 5 assumptions about CLIP's cross-modal transfer and the OG mechanism. This is an empirical ML paper, so the 'axioms' are domain assumptions about model behavior rather than mathematical postulates; the key risk is that the OG explanation is an assumption, not a result.

free parameters (5)
  • Learnable prompt tokens for normal/abnormal per layer = 3 layers x 2 prompts x 12 tokens (CLIP text embedding dim)
    Learned on normal training data via L1_c/L2_c. The prompts define the 'normality' anchor, so the result depends on their fitted values.
  • Softmax temperature tau = 0.001
    Hand-set in Eq. (4); extremely low temperature approximates a one-hot softmax and is not justified by scaling analysis.
  • Constraint loss weight gamma = 0.1
    Hand-set in Eq. (11); controls strength of decoded-feature alignment to normal text.
  • Epoch threshold theta = 5
    Hand-set; L2_c is switched on only after 5 epochs, creating a curriculum whose effect is not analyzed.
  • MoE expert count T and top-K = T=5, K=2
    Selected by scanning on MVTec AD (Table 4); no theory, and same dataset is used for final results.
assumptions (5)
  • domain assumption CLIP text and visual features are aligned enough that generic normal/damaged prompts give discriminative cross-modal responses for industrial patches.
    The whole cross-modal constraint (Eqs. 4-10) depends on this; only an illustrative figure is given, with no quantitative verification.
  • ad hoc to paper Multi-class training degrades performance because the decoder over-generalizes to abnormal patches ('OG'), and text alignment specifically counteracts this.
    Introduction and Section 3.2 state this attribution without direct measurement of abnormal reconstruction.
  • domain assumption A single class-agnostic 'normal object' prompt can capture common textual normality across diverse classes.
    Prompts are shared across all classes; no analysis shows that one prompt serves capsules, PCBs, and foods alike.
  • domain assumption Frozen CLIP encoder layers are suitable for reverse distillation with a decoder of three residual attention blocks.
    Borrowed from RD4AD, but re-applied to CLIP features; the compatibility is assumed, not proven.
  • ad hoc to paper Gaussian noise on encoded features helps mitigate OG, and its magnitude is irrelevant or set implicitly.
    Mentioned in Section 3.1; neither the noise scale nor an ablation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection." pith.science (2026). https://pith.science/paper/NOS62LMD

@misc{pith2026250100346,
  author       = {Pith},
  title        = {Pith review of: CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NOS62LMD}},
  note         = {Machine review of arXiv:2501.00346}
}
read the original abstract

Existing unsupervised distillation-based methods rely on the differences between encoded and decoded features to locate abnormal regions in test images. However, the decoder trained only on normal samples still reconstructs abnormal patch features well, degrading performance. This issue is particularly pronounced in unsupervised multi-class anomaly detection tasks. We attribute this behavior to over-generalization(OG) of decoder: the significantly increasing diversity of patch patterns in multi-class training enhances the model generalization on normal patches, but also inadvertently broadens its generalization to abnormal patches. To mitigate OG, we propose a novel approach that leverages class-agnostic learnable prompts to capture common textual normality across various visual patterns, and then apply them to guide the decoded features towards a normal textual representation, suppressing over-generalization of the decoder on abnormal patterns. To further improve performance, we also introduce a gated mixture-of-experts module to specialize in handling diverse patch patterns and reduce mutual interference between them in multi-class training. Our method achieves competitive performance on the MVTec AD and VisA datasets, demonstrating its effectiveness.

Figures

Figures reproduced from arXiv: 2501.00346 by the authors.

Figure 1
Figure 1. (A) and (B) show the correspondence between visual and text modality and the motivation of CNC, respectively. (C) [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed cross-modal normality distillation framework. Additionally, details of feature-level nor [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization for detection results of UniAD and [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Choices on four pre-trained teacher network [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 39 canonical work pages

  1. [1]

    Bae, J.; Lee, J.-H.; and Kim, S. 2023. PNI: Industrial anomaly detection using position and neighborhood information. In ICCV

    work page 2023

  2. [2]

    Bengio, E.; Bacon, P.-L.; Pineau, J.; and Precup, D. 2015. Conditional computation in neural networks for faster models. arXiv preprint arXiv:1511.06297

    arXiv 2015

  3. [3]

    Bergmann, P.; Fauser, M.; Sattlegger, D.; and Steger, C. 2019. MVTec AD--A comprehensive real-world dataset for unsupervised anomaly detection. In CVPR

    work page 2019

  4. [4]

    Bergmann, P.; Fauser, M.; Sattlegger, D.; and Steger, C. 2020. Uninformed students: Student-teacher anomaly detection with discriminative latent embeddings. In CVPR

    work page 2020

  5. [5]

    Cao, T.; Zhu, J.; and Pang, G. 2023. Anomaly detection under distribution shift. In ICCV

    work page 2023

  6. [6]

    Chen, Y.; Tian, Y.; Pang, G.; and Carneiro, G. 2022. Deep one-class classification via interpolated gaussian descriptor. In AAAI

    work page 2022

  7. [7]

    Z.; Lisanti, G.; and Di Stefano, L

    Costanzino, A.; Ramirez, P. Z.; Lisanti, G.; and Di Stefano, L. 2024. Multimodal industrial anomaly detection by crossmodal feature mapping. In CVPR

    work page 2024

  8. [8]

    Defard, T.; Setkov, A.; Loesch, A.; and Audigier, R. 2021. Padim: A patch distribution modeling framework for anomaly detection and localization. In ICPR

    work page 2021

Show all 52 references
  1. [9]

    Deng, H.; and Li, X. 2022. Anomaly detection via reverse distillation from one-class embedding. In CVPR

    2022

  2. [10]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

    2020 arXiv

  3. [11]

    Fang, Z.; Wang, X.; Li, H.; Liu, J.; Hu, Q.; and Xiao, J. 2023. Fastrecon: Few-shot industrial anomaly detection via fast feature reconstruction. In ICCV

    2023

  4. [12]

    Gu, Z.; Liu, L.; Chen, X.; Yi, R.; Zhang, J.; Wang, Y.; Wang, C.; Shu, A.; Jiang, G.; and Ma, L. 2023. Remembering normality: Memory-guided knowledge distillation for unsupervised anomaly detection. In CVPR

    2023

  5. [13]

    Gu, Z.; Zhang, J.; Liu, L.; Chen, X.; Peng, J.; Gan, Z.; Jiang, G.; Shu, A.; Wang, Y.; and Ma, L. 2024. Rethinking Reverse Distillation for Multi-Modal Anomaly Detection. In AAAI

    2024

  6. [14]

    Gudovskiy, D.; Ishizaka, S.; and Kozuka, K. 2022. Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows. In WACV

    2022

  7. [15]

    Guo, J.; Jia, L.; Zhang, W.; Li, H.; et al. 2024. Recontrast: Domain-specific anomaly detection via contrastive reconstruction. In NeurIPS

    2024

  8. [16]

    He, H.; Bai, Y.; Zhang, J.; He, Q.; Chen, H.; Gan, Z.; Wang, C.; Li, X.; Tian, G.; and Xie, L. 2024 a . Mambaad: Exploring state space models for multi-class unsupervised anomaly detection. arXiv preprint arXiv:2404.06564

    2024 arXiv

  9. [17]

    He, H.; Zhang, J.; Chen, H.; Chen, X.; Li, Z.; Chen, X.; Wang, Y.; Wang, C.; and Xie, L. 2024 b . A diffusion-based framework for multi-class anomaly detection. In AAAI

    2024

  10. [18]

    Hu, T.; Zhang, J.; Yi, R.; Du, Y.; Chen, X.; Liu, L.; Wang, Y.; and Wang, C. 2024. Anomalydiffusion: Few-shot anomaly image generation with diffusion model. In AAAI

    2024

  11. [19]

    Huang, C.; Guan, H.; Jiang, A.; Zhang, Y.; Spratling, M.; and Wang, Y.-F. 2022. Registration based few-shot anomaly detection. In ECCV

    2022

  12. [20]

    Jeong, J.; Zou, Y.; Kim, T.; Zhang, D.; Ravichandran, A.; and Dabeer, O. 2023. Winclip: Zero-/few-shot anomaly classification and segmentation. In CVPR

    2023

  13. [21]

    Jia, M.; Tang, L.; Chen, B.-C.; Cardie, C.; Belongie, S.; Hariharan, B.; and Lim, S.-N. 2022. Visual prompt tuning. In ECCV

    2022

  14. [22]

    Jiang, X.; Liu, J.; Wang, J.; Nie, Q.; Wu, K.; Liu, Y.; Wang, C.; and Zheng, F. 2022. Softpatch: Unsupervised anomaly detection with noisy data. In NeurIPS

    2022

  15. [23]

    Lee, M.; and Choi, J. 2024. Text-guided variational image generation for industrial anomaly detection and segmentation. In CVPR

    2024

  16. [24]

    Lei, J.; Hu, X.; Wang, Y.; and Liu, D. 2023. Pyramidflow: High-resolution defect contrastive localization using pyramid normalizing flow. In CVPR

    2023

  17. [25]

    Li, C.-L.; Sohn, K.; Yoon, J.; and Pfister, T. 2021. Cutpaste: Self-supervised learning for anomaly detection and localization. In CVPR

    2021

  18. [26]

    Li, W.; Xu, X.; Gu, Y.; Zheng, B.; Gao, S.; and Wu, Y. 2024 a . Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Network. In CVPR

    2024

  19. [27]

    Li, Y.; Goodge, A.; Liu, F.; and Foo, C.-S. 2024 b . PromptAD: Zero-shot anomaly detection using text prompts. In CVPR

    2024

  20. [28]

    Lin, J.; and Yan, Y. 2024. A Comprehensive Augmentation Framework for Anomaly Detection. In AAAI

    2024

  21. [29]

    Liu, J.; Xie, G.; Chen, R.; Li, X.; Wang, J.; Liu, Y.; Wang, C.; and Zheng, F. 2024. Real3d-ad: A dataset of point cloud anomaly detection. In NeurIPS

    2024

  22. [30]

    Liu, W.; Chang, H.; Ma, B.; Shan, S.; and Chen, X. 2023 a . Diversity-measurable anomaly detection. In CVPR

    2023

  23. [31]

    Liu, Z.; Zhou, Y.; Xu, Y.; and Wang, Z. 2023 b . Simplenet: A simple network for image anomaly detection and localization. In CVPR

    2023

  24. [32]

    Lu, F.; Yao, X.; Fu, C.-W.; and Jia, J. 2023 a . Removing anomalies as noises for industrial defect localization. In ICCV

    2023

  25. [33]

    Lu, R.; Wu, Y.; Tian, L.; Wang, D.; Chen, B.; Liu, X.; and Hu, R. 2023 b . Hierarchical vector quantized transformer for multi-class unsupervised anomaly detection. In NeurIPS

    2023

  26. [34]

    Ma, J.; Zhao, Z.; Yi, X.; Chen, J.; Hong, L.; and Chi, E. H. 2018. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. In KDD

    2018

  27. [35]

    McIntosh, D.; and Albu, A. B. 2023. Inter-realization channels: Unsupervised anomaly detection beyond one-class classification. In CVPR

    2023

  28. [36]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In ICML

    2021

  29. [37]

    Roth, K.; Pemula, L.; Zepeda, J.; Sch \"o lkopf, B.; Brox, T.; and Gehler, P. 2022. Towards total recall in industrial anomaly detection. In CVPR

    2022

  30. [38]

    Shazeer, N.; Mirhoseini, A.; Maziarz, K.; Davis, A.; Le, Q.; Hinton, G.; and Dean, J. 2017. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv preprint arXiv:1701.06538

    2017 arXiv

  31. [39]

    D.; Nguyen, A

    Tien, T. D.; Nguyen, A. T.; Tran, N. H.; Huy, T. D.; Duong, S.; Nguyen, C. D. T.; and Truong, S. Q. 2023. Revisiting reverse distillation for anomaly detection. In CVPR

    2023

  32. [40]

    Wang, R.; Hoppe, S.; Monari, E.; and Huber, M. F. 2023. Defect transfer gan: Diverse defect synthesis for data augmentation. arXiv preprint arXiv:2302.08366

    2023 arXiv

  33. [41]

    Yao, X.; Li, R.; Zhang, J.; Sun, J.; and Zhang, C. 2023. Explicit boundary guided semi-push-pull contrastive learning for supervised anomaly detection. In CVPR

    2023

  34. [42]

    Yi, J.; and Yoon, S. 2020. Patch svdd: Patch-level svdd for anomaly detection and segmentation. In ACCV

    2020

  35. [43]

    You, Z.; Cui, L.; Shen, Y.; Yang, K.; Lu, X.; Zheng, Y.; and Le, X. 2022. A unified model for multi-class anomaly detection. In NeurIPS

    2022

  36. [44]

    Yu, J.; Zheng, Y.; Wang, X.; Li, W.; Wu, Y.; Zhao, R.; and Wu, L. 2021. Fastflow: Unsupervised anomaly detection and localization via 2d normalizing flows. arXiv preprint arXiv:2111.07677

    2021 arXiv

  37. [45]

    Zavrtanik, V.; Kristan, M.; and Sko c aj, D. 2021. Draem-a discriminatively trained reconstruction embedding for surface anomaly detection. In ICCV

    2021

  38. [46]

    Zhang, J.; Chen, X.; Wang, Y.; Wang, C.; Liu, Y.; Li, X.; Yang, M.-H.; and Tao, D. 2023. Exploring plain vit reconstruction for multi-class unsupervised anomaly detection. arXiv preprint arXiv:2312.07495

    2023 arXiv

  39. [47]

    Zhang, X.; Xu, M.; and Zhou, X. 2024. RealNet: A feature selection network with realistic synthetic anomaly for anomaly detection. In CVPR

    2024

  40. [48]

    C.; and Liu, Z

    Zhou, K.; Yang, J.; Loy, C. C.; and Liu, Z. 2022. Learning to prompt for vision-language models. IJCV, 130(9): 2337--2348

    2022

  41. [49]

    Zhou, Q.; Pang, G.; Tian, Y.; He, S.; and Chen, J. 2023. Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection. arXiv preprint arXiv:2310.18961

    2023

  42. [50]

    Zou, Y.; Jeong, J.; Pemula, L.; Zhang, D.; and Dabeer, O. 2022. Spot-the-difference self-supervised pre-training for anomaly detection and segmentation. In ECCV

    2022

  43. [51]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  44. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.