Pith. sign in

REVIEW 5 major objections 7 minor 1 cited by

Memory-Augmented Dual-Decoder Networks for Multi-Class Unsupervised Anomaly Detection

T0 review · 5 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A dual-decoder network detects anomalies across 15 classes at once, lifting pixel mAP on MVTec AD from 55.4 to 66.4.

desk verdict Solid engineering result with large reported gains on four benchmarks, but the RID mechanism story is under-supported and the identity decoder has a genuine training/inference inconsistency that the authors need to fix. read the letter →

arxiv 2504.14884 v1 pith:OQZKZJR2 submitted 2025-04-21 cs.CV

Jingyu Xing , Chenwei Tang , Tao Wang , Rong Xiao , Wei Ju , Ji-Zhe Zhou , Liangli Zhen , Jiancheng Lv This is my paper · ORCID
classification cs.CV
keywords multi-classanomalydetectionunsupervisedfeaturereconstructionreversedistillationclass-awarememorydualdecoderlocalization
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

Reconstruction-based anomaly detection faces two opposite failure modes in multi-class settings: subtle or compositionally similar anomalies are reconstructed too well and escape detection, while intricate normal textures are reconstructed too poorly and produce false alarms. The paper proposes MDD-Net, which attacks both at once by training two structurally identical decoders side by side: a restoration decoder that turns synthetic anomalies back into normal features, and an identity decoder that deliberately preserves anomalous semantics. Because the two decoders share architecture and capacity, the paper argues their errors on hard normal patterns are correlated and cancel when the outputs are compared, leaving the Restoration-Identity Discrepancy as a cleaner anomaly signal. A class-aware memory module replaces anomalous tokens with learned normal prototypes, suppressing anomaly reconstruction directly. On MVTec AD the method reports image-level AUROC of 98.6 and pixel-level mAP of 66.4 against 55.4 for ViTAD, with similar gains on VisA, Real-IAD, and Uni-Medical.

What carries the argument

The load-bearing object is the Restoration-Identity Discrepancy (RID), the summed cosine distance between the multi-stage feature maps produced by two identically structured ViT decoders. The Restoration Decoder receives teacher features after the class-aware memory module replaces each token with a sparse combination of learned normal prototypes, so it is trained to reconstruct normal features from anomaly-contaminated input; the Identity Decoder receives teacher features without that restoration objective and is trained to reproduce the anomalous semantics. The paper's argument is that because both decoders have identical structure and comparable capacity, their representational deficiencies are correlated, so on normal regions the two outputs fail in similar ways and the difference cancels, while on anomaly regions the restoration branch moves toward normality and the identity branch stays with the anomaly, producing a large discrepancy. The Class-aware Memory Module carries the second half of the mechanism: sparse cosine retrieval from a bank of class-labeled prototypes, a per-token classification loss that disentangles class-specific from shared prototypes, and gradient truncation that keeps anomaly features from contaminating the memory.

What would settle it

On a validation set of only normal images, compute per-pixel cosine distance maps between the restoration and identity decoders and measure the spatial correlation of their individual reconstruction errors; if the two error maps are weakly correlated, the RID signal cannot cancel normal-region false positives. As a second check, retrain with the identity decoder using an architecturally different decoder; if performance does not drop, the correlated-deficiency assumption is not load-bearing.

Watch

Extended reading notes

Core claim

The paper's central claim is that over-generalization and insufficient normality reconstruction are not a trade-off but two sides of the same decoder, and both can be corrected by decoupling what the decoder is asked to do. The restoration branch is trained to output normal teacher features after the memory module has stripped anomaly content from the input tokens; the identity branch is trained to preserve the anomaly semantics present in the teacher features. At inference, the pixel-wise cosine distance between the two branches, fused with the teacher-restoration distance, forms the anomaly map. The paper reports that this scheme improves multi-class anomaly detection and localization on all four benchmarks, with the largest margins at the pixel level: MVTec AD pixel mAP of 66.4 versus 55.4 for ViTAD, pixel mIoU of 48.8 versus 42.5, and a 3.6-point average image-level gain over the previous best method.

Load-bearing premise

Everything rests on the claim that two structurally identical decoders with comparable capacity fail in correlated ways on hard normal patterns; if their errors do not co-occur, subtracting their outputs creates false positives on normal textures instead of removing them.

Editorial extensions

If this is right

  • A single trained model can replace per-class detectors in multi-class industrial inspection: on MVTec AD, one MDD-Net covers 15 classes with 98.6 image-level AUROC and 66.4 pixel mAP.
  • The largest reported gains are in localization, so the method is suited to tasks that need pixel-accurate anomaly maps, such as generating weak supervision for segmentation.
  • The same architecture transfers from industrial textures and objects to heterogeneous medical modalities without per-class retraining, with pixel mAP rising to 57.8 on Uni-Medical versus 48.2 for ViTAD.
  • Because the class-aware memory learns both class-specific and shared prototypes, class labels during training are part of the method, and the authors identify removing that label dependency as the main limitation.
  • The two discrepancy signals can be fused by simple linear combination and the losses summed directly, suggesting the method does not require delicate multi-loss tuning.

Reading between the lines

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

  • An implicit consequence of the correlated-deficiency argument is that the same dual-decoder trick should transfer to other dense reconstruction tasks where a student model has systematic blind spots; the paper does not test this.
  • A controlled experiment that feeds both decoders the same transformed input at training and inference has not been reported; the pseudocode routes the identity branch differently between the two phases, so the RID signal's dependence on that pipeline difference is untested.
  • The class-aware memory's visualization suggests cross-class texture prototypes are shared (leather and grid); a label-free clustering variant would clarify how much of the gain comes from class supervision versus from memory replacement alone.
  • Because the main reported wins are pixel-level, evaluating the method on anomaly benchmarks with finer-grained ground-truth masks beyond the four used here would show whether the localization gain generalizes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 7 minor

Summary. The paper proposes MDD-Net, a multi-class unsupervised anomaly detection framework that combines a dual-decoder reverse distillation network (DRD-Net) with a class-aware memory module (CMM). A frozen teacher encoder extracts features from normal and synthetic anomalous images; the restoration decoder reconstructs normal features from memory-replaced anomaly tokens, while the identity decoder is intended to preserve anomaly semantics. Anomaly maps are formed by fusing a Restoration-Identity Discrepancy (RID) and a Teacher-Restoration Discrepancy (TRD). The method is evaluated on MVTec AD, VisA, Real-IAD, and Uni-Medical under eight image- and pixel-level metrics, with ablations on MVTec AD. The central claims are that MDD-Net simultaneously mitigates over-generalization and insufficient normal reconstruction, achieving a 3.6% average improvement over prior state-of-the-art on MVTec AD and large pixel-level gains on all benchmarks.

Significance. If the mechanism and the empirical claims are sound, MDD-Net would be a useful step toward a single unified model for multi-class anomaly detection and localization, directly addressing the named failure modes of over-generalization and insufficient normality reconstruction. The paper's strengths include its breadth of evaluation (four benchmarks and eight metrics) and the explicit component-wise ablations in Tables 3 and 4. However, the central mechanism for false-positive reduction is supported by an asserted rather than measured correlation between the two decoders, and the training/inference protocol for the identity decoder contains a concrete mismatch. The reported 'superior over SoTA' claim is also not uniformly supported by the numbers in Table 1, particularly on Real-IAD. These issues are load-bearing for the paper's main conclusions.

major comments (5)
  1. [Section 3.2, Algorithm 1, Figure 2] The identity decoder is trained on raw stage-4 tokens F_a^4_T (Algorithm 1 Phase 2 step 4 and Eq. 3), but at inference it receives Neck(F_T^4) (Phase 3 step 2). Figure 2's step 5 caption states that anomalous features are routed through Neck and Identity Decoder during training, which contradicts both the main text and Algorithm 1. This mismatch is load-bearing because the paper's central claim (Section 3.2, Section 3.5) is that RID cancels errors on normal regions thanks to the correlated deficiencies of two structurally identical decoders. When the identity decoder sees a different input transform at test time, and when it has never seen normal-only images during training, the correlated-error-cancellation argument does not apply; RID can become large on normal pixels. Please standardize the input pathway between training and inference and report RID statistics separately for normal pixels and anomalous pixels on real test images.
  2. [Section 3.2, Table 3, Eq. 5] The correlated-capacity argument is asserted rather than measured. The large improvement from DRD-Net+L over DRD-Net-L in Table 3 (P-mAP 61.91 vs 56.73, P-mIoU 44.90 vs 43.64) may be driven mainly by L_dist, which directly supervises M_RI with the synthetic anomaly mask via Eq. 5, rather than by emergent cancellation of correlated decoder errors on normal patterns. To separate these explanations, please add an ablation that removes L_dist while retaining the identity decoder, and directly measure decoder error correlation on hard-normal patches, for example by computing per-pixel TRD and RID on normal-only test images and reporting their joint distribution.
  3. [Section 4.1, Implementation Details] The fusion ratio alpha (0.4, 0.4, 0.1, 0.5 for the four datasets) and memory bank size N (500, 1000, 2000) are reported without a validation protocol. Because alpha directly controls how much weight the proposed RID term receives, tuning it per benchmark on the test set can inflate the apparent gains of the method. The paper should either fix these hyperparameters across datasets, justify them on a held-out validation split, or report sensitivity curves showing that the conclusions are robust to the chosen values. In addition, no repeated-run statistics or error bars are provided, and some reported gains are small enough to fall within typical run-to-run variation.
  4. [Table 1 and Section 4.2] Table 1 does not support the statement that MDD-Net is superior to current state-of-the-art approaches across all four benchmarks. On Real-IAD, MambaAD achieves 87.0 mAUROC, 85.3 image-level mAP, 98.6 max, 32.4 pixel mAUROC, 38.1 pixel mAP, 91.2 pixel mF1, and 23.9 mAU-PRO, while MDD-Net achieves 86.7, 85.1, 97.7, 30.8, 35.9, 86.5, and 22.3, respectively; MDD-Net is worse on every listed Real-IAD metric except a tie on mF1-max (77.6). On VisA, MDD-Net is also below MambaAD on image-level AUROC and pixel-level mF1. The abstract and introduction should be revised to present per-dataset comparisons accurately rather than claiming uniform SoTA performance.
  5. [Section 3.3, Eq. 11, Limitations] The method is described as 'unsupervised' in the title and abstract, but the CMM training uses class labels y through the classification loss L_cls in Eq. 11. This contradicts the Introduction's statement that training uses 'only normal samples.' The Limitations paragraph acknowledges reliance on category labels, but the framing is still misleading. Please reclassify the setting as label-guided or weakly supervised multi-class anomaly detection, or provide a clear justification for why the availability of class labels during training is compatible with the unsupervised claim.
minor comments (7)
  1. [Eq. (11)] The inner summation is written as 'D=1' to 'D' with the dummy index D colliding with the dimension D; please change the dummy index to c, i.e., sum_{c=1}^{D} exp(\hat{y}^{(t)}_c).
  2. [References] References [1] and [2] are identical duplicates (both cite the same BMAD paper); one should be removed or replaced with the intended second citation.
  3. [Algorithm 1, Supplementary] In the L_rec term of Algorithm 1, the parentheses are mismatched: 'Dcos(f_R(Neck(M(F_n^4_T)), F_n^i_T)' should be 'Dcos(f_R(Neck(M(F_n^4_T))), F_n^i_T)'. Also, the notation Dcos is used in the supplement but not defined in the main text.
  4. [Section 4.2] The sentence 'MDD-Net segments anonymous areas more accurately' should read 'anomalous areas.'
  5. [Figure 2 caption] The caption for step 5 says the anomalous features are reconstructed through Neck and Identity Decoder, which conflicts with Section 3.2 and Algorithm 1 (where the identity decoder receives raw F_a^4_T). Please align these descriptions.
  6. [Section 4.1] The sentence 'Divergences across datasets exclusively manifest in memory bank capacity and batch dimension configurations' is inaccurate because the fusion ratio alpha also differs per dataset; please correct it.
  7. [Abstract] The abstract states an 'average improvement of 3.6%' over state-of-the-art on MVTec AD; please specify the baseline and the set of metrics used to compute this average, since Table 1 shows different gains for different metrics.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the central claims rest on held-out test evaluations, with only minor non-circular tuning concerns.

full rationale

The paper's central derivation is not circular. The two decoders are trained with explicit cosine losses (Eqs. 2-3) and the discrepancy map is supervised against synthetic masks (Eqs. 4-5), but this is a standard pseudo-anomaly training signal: the reported test results are on held-out real images of MVTec AD, VisA, Real-IAD, and Uni-Medical, so the prediction does not reduce by construction to the training masks. The memory module and prototype replacement are trained with reconstruction and classification losses that do not encode the test labels. No load-bearing claim is justified solely by a self-citation; references to MemAE, DRAEM, and DeSTSeg are external baselines. Minor non-circular concerns: the fusion ratio alpha is reported as per-dataset 'optimal' (Section 4.1), which is a hyperparameter-tuning leak rather than a circular reduction, and the identity decoder is trained on raw F_a^4_T but fed Neck(F_T^4) at inference (Algorithm 1), an input-distribution mismatch that threatens the RID mechanism but does not make the derivation circular. Overall the empirical claims rest on independent test-set evaluation, so the circularity score is low.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The method is an empirical combination whose main burden is the transfer of synthetic anomalies to real defects, the correlated-capacity assumption for the two decoders, and the availability of global class labels. The central reported gains also depend on per-dataset tuned hyperparameters such as alpha and memory size.

free parameters (4)
  • Memory bank size N = 500 (MVTec AD), 1000 (VisA, Uni-Medical), 2000 (Real-IAD)
    Chosen per dataset; capacity changes retrieval sparsity and affects performance, as shown by the sensitivity analysis in Supplementary Figure 2.
  • Fusion ratio alpha for RID/TRD = 0.4 (MVTec AD), 0.4 (VisA), 0.1 (Real-IAD), 0.5 (Uni-Medical)
    Controls the final anomaly map S = alpha * S_RI + (1-alpha) * S_TR; tuned per dataset and directly affects reported scores.
  • Hard shrinkage threshold lambda = 1/N for all datasets
    Controls sparsity of memory retrieval; the paper notes a typical range of [1/N, 3/N] and fixes lambda = 1/N.
  • Online hard example mining cosine threshold = not reported
    Gradients are discarded for well-reconstructed feature points exhibiting low cosine distances; the threshold is not quantified, which makes exact reproduction impossible.
assumptions (5)
  • domain assumption Synthetic anomalies produced by DRAEM (Perlin noise plus DTD textures) are a faithful proxy for real anomalies
    All decoder training, L_dist mask supervision, and restoration/identity behavior are learned on these synthetic pairs in Section 3.2. If real defects differ from synthetic ones, the learned RID/TRD signal may not transfer.
  • ad hoc to paper Two structurally identical decoders with comparable capacity have correlated representational deficiencies on hard-normal features
    This argument in Section 3.2 is the basis for using Restoration-Identity Discrepancy to cancel false positives; no analysis or experiment directly validates the correlation.
  • domain assumption Frozen DINO-ViT features provide a sufficiently discriminative and stable normality representation for multi-class reconstruction
    The teacher encoder is fixed with DINO-S weights and all losses are computed in its feature space, as described in Section 3.1. The method inherits the quality of these features.
  • domain assumption Global class labels are available for all normal training images
    L_cls in Eq. 11 supervises token-level memory retrieval with the image class label; the authors acknowledge in the Limitations that this is a dependence on category labels.
  • domain assumption Sparse memory retrieval prevents anomalous features from being reconstructed from normal prototypes
    CMM replaces tokens by sparse combinations of memory prototypes in Section 3.3; if the prototype bank is too coarse or contaminated, anomalies may still be approximated.
invented entities (1)
  • Class-aware memory prototypes with learnable per-item class probability vectors
    purpose: Store class-specific and shared normal prototypes and steer restoration away from anomaly-like features, as described in Section 3.3.
    Evidence is internal: t-SNE, silhouette score, Davies-Bouldin index on the model's own memory in Figure 5, and ablations in Table 4. No external benchmark isolates the class-probability mechanism.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Memory-Augmented Dual-Decoder Networks for Multi-Class Unsupervised Anomaly Detection." pith.science (2026). https://pith.science/paper/OQZKZJR2

@misc{pith2026250414884,
  author       = {Pith},
  title        = {Pith review of: Memory-Augmented Dual-Decoder Networks for Multi-Class Unsupervised Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OQZKZJR2}},
  note         = {Machine review of arXiv:2504.14884}
}
read the original abstract

Recent advances in unsupervised anomaly detection (UAD) have shifted from single-class to multi-class scenarios. In such complex contexts, the increasing pattern diversity has brought two challenges to reconstruction-based approaches: (1) over-generalization: anomalies that are subtle or share compositional similarities with normal patterns may be reconstructed with high fidelity, making them difficult to distinguish from normal instances; and (2) insufficient normality reconstruction: complex normal features, such as intricate textures or fine-grained structures, may not be faithfully reconstructed due to the model's limited representational capacity, resulting in false positives. Existing methods typically focus on addressing the former, which unintentionally exacerbate the latter, resulting in inadequate representation of intricate normal patterns. To concurrently address these two challenges, we propose a Memory-augmented Dual-Decoder Networks (MDD-Net). This network includes two critical components: a Dual-Decoder Reverse Distillation Network (DRD-Net) and a Class-aware Memory Module (CMM). Specifically, the DRD-Net incorporates a restoration decoder designed to recover normal features from synthetic abnormal inputs and an identity decoder to reconstruct features that maintain the anomalous semantics. By exploiting the discrepancy between features produced by two decoders, our approach refines anomaly scores beyond the conventional encoder-decoder comparison paradigm, effectively reducing false positives and enhancing localization accuracy. Furthermore, the CMM explicitly encodes and preserves class-specific normal prototypes, actively steering the network away from anomaly reconstruction. Comprehensive experimental results across several benchmarks demonstrate the superior performance of our MDD-Net framework over current SoTA approaches in multi-class UAD tasks.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. In-batch Relational Features Enhance Precision in An Unsupervised Medical Anomaly Detection Task

    q-bio.QM 2026-03 conditional novelty 6.0 of 10

    Augmenting an autoencoder's latent code with k-nearest-neighbor context from the same batch improves unsupervised brain-MRI anomaly detection (AUC 0.90, AP 0.78 vs 0.84/0.62 baseline).

Reference graph

Works this paper leans on

53 extracted references · 5 linked inside Pith · cited by 1 Pith paper

  1. [2]

    Jinan Bao, Hanshi Sun, Hanqiu Deng, Yinsheng He, Zhaoxiang Zhang, and Xingyu Li. 2024. Bmad: Benchmarks for medical anomaly detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 4042–4053

    2024

  2. [3]

    Paul Bergmann, Kilian Batzner, Michael Fauser, David Sattlegger, and Carsten Steger. 2022. Beyond dents and scratches: Logical constraints in unsupervised anomaly detection and localization. International Journal of Computer Vision 130, 4 (2022), 947–969

    2022

  3. [4]

    Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. 2019. MVTec AD–A comprehensive real-world dataset for unsupervised anomaly de- tection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9592–9600

    2019

  4. [5]

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. 2021. Emerging properties in self-supervised vision transformers. In Proceedings of the IEEE/CVF international conference on computer vision. 9650–9660

    2021

  5. [6]

    Qiyu Chen, Huiyuan Luo, Chengkan Lv, and Zhengtao Zhang. 2024. A unified anomaly synthesis strategy with gradient ascent for industrial anomaly detection and localization. In European Conference on Computer Vision . Springer, 37–54

    2024

  6. [7]

    Zhang Chen, Zhiqiang Tian, Jihua Zhu, Ce Li, and Shaoyi Du. 2022. C-cam: Causal cam for weakly supervised semantic segmentation on medical image. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 11676–11685

    2022

  7. [8]

    Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and An- drea Vedaldi. 2014. Describing textures in the wild. In Proceedings of the IEEE conference on computer vision and pattern recognition . 3606–3613

    2014

  8. [9]

    David L Davies and Donald W Bouldin. 1979. A cluster separation measure. IEEE transactions on pattern analysis and machine intelligence 2 (1979), 224–227

    1979

Show all 53 references
  1. [10]

    Thomas Defard, Aleksandr Setkov, Angelique Loesch, and Romaric Audigier

  2. [11]

    Hanqiu Deng and Xingyu Li. 2022. Anomaly detection via reverse distillation from one-class embedding. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 9737–9746

    2022

  3. [12]

    Dong Gong, Lingqiao Liu, Vuong Le, Budhaditya Saha, Moussa Reda Mansour, Svetha Venkatesh, and Anton van den Hengel. 2019. Memorizing normality to detect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection. In Proceedings of the IEEE/CVF international...

    2019

  4. [13]

    Zhihao Gu, Liang Liu, Xu Chen, Ran Yi, Jiangning Zhang, Yabiao Wang, Chengjie Wang, Annan Shu, Guannan Jiang, and Lizhuang Ma. 2023. Remembering normal- ity: Memory-guided knowledge distillation for unsupervised anomaly detection. In Proceedings of the IEEE/CVF International C...

    2023

  5. [14]

    Denis Gudovskiy, Shun Ishizaka, and Kazuki Kozuka. 2022. Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows. In Proceedings of the IEEE/CVF winter conference on applications of computer vision. 98–107

    2022

  6. [15]

    Haoyang He, Yuhu Bai, Jiangning Zhang, Qingdong He, Hongxu Chen, Zhenye Gan, Chengjie Wang, Xiangtai Li, Guanzhong Tian, and Lei Xie. 2024. Mam- baAD: Exploring State Space Models for Multi-class Unsupervised Anom- aly Detection. In Advances in Neural Information Processing Sy...

    2024

  7. [16]

    Haoyang He, Jiangning Zhang, Hongxu Chen, Xuhai Chen, Zhishan Li, Xu Chen, Yabiao Wang, Chengjie Wang, and Lei Xie. 2024. A diffusion-based framework for multi-class anomaly detection. In Proceedings of the AAAI conference on artificial intelligence, Vol. 38. 8472–8480

    2024

  8. [17]

    Liren He, Zhengkai Jiang, Jinlong Peng, Wenbing Zhu, Liang Liu, Qiangang Du, Xiaobin Hu, Mingmin Chi, Yabiao Wang, and Chengjie Wang. 2024. Learning Uni- fied Reference Representation for Unsupervised Multi-class Anomaly Detection. In European Conference on Computer Vision . S...

    2024

  9. [18]

    Jiulin Lang, Chenwei Tang, Yi Gao, and Jiancheng Lv. 2021. Knowledge Distilla- tion Method for Surface Defect Detection. InNeural Information Processing, Teddy Mantoro, Minho Lee, Media Anugerah Ayu, Kok Wai Wong, and Achmad Nizar Hidayanto (Eds.). Springer International Publi...

    2021

  10. [19]

    Chun-Liang Li, Kihyuk Sohn, Jinsung Yoon, and Tomas Pfister. 2021. Cutpaste: Self-supervised learning for anomaly detection and localization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 9664–9674

    2021

  11. [20]

    Yufei Liang, Jiangning Zhang, Shiwei Zhao, Runze Wu, Yong Liu, and Shuwen Pan. 2023. Omni-frequency channel-selection representations for unsupervised anomaly detection. IEEE Transactions on Image Processing 32 (2023), 4327–4340

    2023

  12. [21]

    Xinyue Liu, Jianyuan Wang, Biao Leng, and Shuo Zhang. 2024. Unlocking the Potential of Reverse Distillation for Anomaly Detection. arXiv preprint arXiv:2412.07579 (2024)

    2024 arXiv

  13. [22]

    Yang Liu, Zhaoyang Xia, Mengyang Zhao, Donglai Wei, Yuzheng Wang, Siao Liu, Bobo Ju, Gaoyun Fang, Jing Liu, and Liang Song. 2023. Learning causality- inspired representation consistency for video anomaly detection. In Proceedings of the 31st ACM international conference on mul...

    2023

  14. [23]

    Zhikang Liu, Yiming Zhou, Yuansheng Xu, and Zilei Wang. 2023. Simplenet: A simple network for image anomaly detection and localization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 20402–20411

    2023

  15. [24]

    Ruiying Lu, YuJie Wu, Long Tian, Dongsheng Wang, Bo Chen, Xiyang Liu, and Ruimin Hu. 2023. Hierarchical vector quantized transformer for multi-class unsupervised anomaly detection. Advances in Neural Information Processing Systems 36 (2023), 8487–8500

    2023

  16. [25]

    Ju-Hyeon Nam, Nur Suriza Syazwany, Su Jung Kim, and Sang-Chul Lee. 2024. Modality-agnostic domain generalizable medical image segmentation by multi- frequency in multi-scale attention. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 11480–11491

    2024

  17. [26]

    Cheng Ouyang, Carlo Biffi, Chen Chen, Turkay Kart, Huaqi Qiu, and Daniel Rueckert. 2022. Self-supervised learning for few-shot medical image segmenta- tion. IEEE Transactions on Medical Imaging 41, 7 (2022), 1837–1848

    2022

  18. [27]

    Guansong Pang, Cheng Yan, Chunhua Shen, Anton van den Hengel, and Xiao Bai. 2020. Self-trained deep ordinal regression for end-to-end video anomaly detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 12173–12182

    2020

  19. [28]

    Jonathan Pirnay and Keng Chai. 2022. Inpainting transformer for anomaly detection. In International Conference on Image Analysis and Processing . Springer, MM ’25, October 27–31, 2025, Dublin, Ireland Trovato et al. 394–406

    2022

  20. [29]

    Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Schölkopf, Thomas Brox, and Peter Gehler. 2022. Towards total recall in industrial anomaly detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 14318–14328

    2022

  21. [30]

    Hannah M Schlüter, Jeremy Tan, Benjamin Hou, and Bernhard Kainz. 2022. Nat- ural synthetic anomalies for self-supervised anomaly detection and localization. In European Conference on Computer Vision . Springer, 474–489

    2022

  22. [31]

    Ketan Rajshekhar Shahapure and Charles Nicholas. 2020. Cluster quality analysis using silhouette score. In 2020 IEEE 7th international conference on data science and advanced analytics (DSAA) . IEEE, 747–748

    2020

  23. [32]

    Chenrui Shi, Che Sun, Yuwei Wu, and Yunde Jia. 2023. Video anomaly detection via sequentially learning multiple pretext tasks. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 10330–10340

    2023

  24. [33]

    Chenwei Tang, Caiyang Yu, Yi Gao, Jianming Chen, Jiaming Yang, Jiuling Lang, Chuan Liu, Ling Zhong, Zhenan He, and Jiancheng Lv. 2022. Deep learning in nuclear industry: A survey. Big Data Mining and Analytics 5, 2 (2022), 140–160. https://doi.org/10.26599/BDMA.2021.9020027

    2022

  25. [34]

    Tran Dinh Tien, Anh Tuan Nguyen, Nguyen Hoang Tran, Ta Duc Huy, Soan Duong, Chanh D Tr Nguyen, and Steven QH Truong. 2023. Revisiting reverse distillation for anomaly detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 24511–24520

    2023

  26. [35]

    Chengjie Wang, Wenbing Zhu, Bin-Bin Gao, Zhenye Gan, Jiangning Zhang, Zhihao Gu, Shuguang Qian, Mingang Chen, and Lizhuang Ma. 2024. Real-iad: A real-world multi-view dataset for benchmarking versatile industrial anomaly detection. In Proceedings of the IEEE/CVF Conference on ...

    2024

  27. [36]

    Xiaolei Wang, Xiaoyang Wang, Huihui Bai, Eng Gee Lim, and Jimin Xiao. 2024. CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection. arXiv preprint arXiv:2501.00346 (2024)

    2024 arXiv

  28. [37]

    Tiange Xiang, Yixiao Zhang, Yongyi Lu, Alan L Yuille, Chaoyi Zhang, Weidong Cai, and Zongwei Zhou. 2023. Squid: Deep feature in-painting for unsupervised anomaly detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 23890–23901

    2023

  29. [38]

    Xiaoli Yao, Jia Tan, Zijian Deng, Deng Xiong, Qijun Zhao, and Min Wu. 2025. MUPO-Net: A Multilevel Dual-domain Progressive Enhancement Network with Embedded Attention for CT Metal Artifact Reduction. In ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Si...

    2025

  30. [39]

    Zhiyuan You, Lei Cui, Yujun Shen, Kai Yang, Xin Lu, Yu Zheng, and Xinyi Le

  31. [40]

    Caiyang Yu, Yixi Wang, Chenwei Tang, Wentao Feng, and Jiancheng Lv. 2023. EU- Net: Automatic U-Net neural architecture search with differential evolutionary algorithm for medical image segmentation. Computers in Biology and Medicine 167 (2023), 107579. https://doi.org/10.1016/...

    2023

  32. [41]

    Vitjan Zavrtanik, Matej Kristan, and Danijel Skočaj. 2021. Draem-a discrimi- natively trained reconstruction embedding for surface anomaly detection. In Proceedings of the IEEE/CVF international conference on computer vision . 8330– 8339

    2021

  33. [42]

    Vitjan Zavrtanik, Matej Kristan, and Danijel Skočaj. 2021. Reconstruction by inpainting for visual anomaly detection. Pattern Recognition 112 (2021), 107706

    2021

  34. [43]

    Fangyuan Zhang, Tianxiang Pan, Jun-Hai Yong, and Bin Wang. 2024. W2P: switching from weak supervision to partial supervision for semantic segmen- tation. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 38. 7042–7050

    2024

  35. [44]

    Hui Zhang, Zheng Wang, Zuxuan Wu, and Yu-Gang Jiang. 2023. DiffusionAD: Norm-guided one-step denoising diffusion for anomaly detection. arXiv preprint arXiv:2303.08730 (2023)

    2023 arXiv

  36. [45]

    Jiangning Zhang, Xuhai Chen, Yabiao Wang, Chengjie Wang, Yong Liu, Xiangtai Li, Ming-Hsuan Yang, and Dacheng Tao. 2025. Exploring plain ViT features for multi-class unsupervised visual anomaly detection. Computer Vision and Image Understanding (2025), 104308

    2025

  37. [46]

    Jiangning Zhang, Haoyang He, Zhenye Gan, Qingdong He, Yuxuan Cai, Zhucun Xue, Yabiao Wang, Chengjie Wang, Lei Xie, and Yong Liu. 2024. A Comprehensive Library for Benchmarking Multi-class Visual Anomaly Detection. arXiv preprint arXiv:2406.03262 (2024)

    2024 arXiv

  38. [47]

    Jiangning Zhang, Chengjie Wang, Xiangtai Li, Guanzhong Tian, Zhucun Xue, Yong Liu, Guansong Pang, and Dacheng Tao. 2024. Learning Feature Inversion for Multi-class Anomaly Detection under General-purpose COCO-AD Benchmark. arXiv preprint arXiv:2404.10760 (2024)

    2024 arXiv

  39. [48]

    Xuan Zhang, Shiyu Li, Xi Li, Ping Huang, Jiulong Shan, and Ting Chen. 2023. Destseg: Segmentation guided denoising student-teacher for anomaly detec- tion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 3914–3923

    2023

  40. [49]

    Ximiao Zhang, Min Xu, and Xiuzhuang Zhou. 2024. Realnet: A feature selection network with realistic synthetic anomaly for anomaly detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition . 16699–16708

    2024

  41. [50]

    Ying Zhao. 2023. Omnial: A unified cnn framework for unsupervised anomaly localization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 3924–3933

    2023

  42. [51]

    Yixuan Zhou, Xing Xu, Jingkuan Song, Fumin Shen, and Heng Tao Shen. 2024. Msflow: Multiscale flow-based framework for unsupervised anomaly detection. IEEE Transactions on Neural Networks and Learning Systems (2024)

    2024

  43. [52]

    Yang Zou, Jongheon Jeong, Latha Pemula, Dongqing Zhang, and Onkar Dabeer

  44. [2021]

    In International conference on pattern recognition

    Padim: a patch distribution modeling framework for anomaly detection and localization. In International conference on pattern recognition . Springer, 475– 489

  45. [2022]

    Advances in Neural Information Processing Systems 35 (2022), 4571–4584

    A unified model for multi-class anomaly detection. Advances in Neural Information Processing Systems 35 (2022), 4571–4584

    2022

Pith tools

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