SuperAD: A Training-free Anomaly Classification and Segmentation Method for CVPR 2025 VAND 3.0 Workshop Challenge Track 1: Adapt & Detect

Hang Chen; Huaiyuan Zhang; Lei Qi; Linao Han; Linghao Sun; Shunyi Wu; Yu Cheng; Zeyu Shi

REVIEW 4 major objections 5 minor 14 references

SuperAD: A Training-free Anomaly Classification and Segmentation Method for CVPR 2025 VAND 3.0 Workshop Challenge Track 1: Adapt & Detect

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A training-free method claims to beat all trained anomaly detectors on MVTec AD 2.

desk verdict Training-free DINOv2 memory-bank method on MVTec AD 2 looks promising, but the huge F1 numbers are undermined by test-set threshold tuning and category-specific post-processing. read the letter →

arxiv 2505.19750 v2 pith:L5YN4HM3 submitted 2025-05-26 cs.CV

Huaiyuan Zhang , Hang Chen , Yu Cheng , Shunyi Wu , Linghao Sun , Linao Han , Zeyu Shi , Lei Qi This is my paper · ORCID

classification cs.CV

keywords anomalydetectionsegmentationMVTecAD2DINOv2memorybanknearestneighbortraining-freeindustrialinspection

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

SuperAD claims that a fully training-free procedure can outperform every trained anomaly-detection baseline on the MVTec AD 2 benchmark, which contains transparent, reflective, and occluded industrial objects under varied lighting. The method extracts features from four layers of the pre-trained DINOv2-large vision transformer, compares each test patch against a small memory bank of 16 normal reference images per category chosen by greedy coreset selection, and averages the per-layer nearest-neighbor distance maps into a segmentation. On the private test set it reports a mean segmentation F1 of 47.8% (43.2% on a mixed-lighting set), compared with 21.8% for the best prior method on the standard set. If correct, this would show that the representational power of large self-supervised models alone, without any task-specific training or fine-tuning, is sufficient to handle the physical complexity and lighting shifts that defeat current trained detectors.

What carries the argument

The central mechanism is a memory-bank nearest-neighbor anomaly score computed from multi-scale DINOv2 features. DINOv2 is a self-supervised vision transformer pre-trained on a large unlabeled image corpus; here its ViT-L-14 variant supplies frozen feature maps at four depths. Greedy coreset selection (the PatchCore procedure) picks 16 normal reference images per category so the bank spans the variety of normal appearances. For a test image, each spatial feature vector is matched to its nearest neighbor in the bank at each of the four layers; the distance maps are averaged and upsampled to a final anomaly map. Two additional hand-crafted steps support the core mechanism: a PCA-based binarization of shallow-layer features produces a foreground mask used for Vial and Wallplugs, and closed-region filling is applied for Fabric and Walnuts.

What would settle it

Run the identical pipeline on the MVTec AD 2 private protocol with the segmentation threshold fixed across all categories at a single value not derived from TESTpublic labels, and with the Fabric/Walnuts closed-region filling disabled; if the mean F1 then falls below the best trained baseline, the claimed advantage rests on public-set calibration and category-specific heuristics rather than the training-free representation.

Watch

Extended reading notes

Core claim

The central discovery claimed is that nearest-neighbor retrieval against a frozen DINOv2 feature bank is a stronger real-world anomaly segmenter than learned reconstruction, distillation, or flow-based methods when the benchmark contains transparent materials, bulk objects, and lighting variation. For each of the eight categories, 16 normal images are selected greedily to maximize diversity; features from layers 6, 12, 18, and 24 of DINOv2-large are stored as the memory bank. At test time, every patch's nearest-neighbor distance to this bank is computed at each layer, and the four distance maps are averaged and upsampled. The paper attributes the method's robustness under lighting shifts to this matching process: normal regions find similar stored features across lighting conditions, while anomalies do not. Reported per-category gains are large (Can: 17.3% vs 5.0% best prior; Rice: 60.9% vs 19.1%), and the paper additionally reports competitive ClassF1 scores of 70.2% and 74.4% for image-level classification on the two private test sets.

Load-bearing premise

The fixed configuration (16 reference images, DINOv2 layers 6/12/18/24, PCA threshold 1.0, 3x3 morphological kernel, and the category-specific foreground-mask and fill-region post-processing, along with an F1 threshold tuned on the public test set) transfers unchanged to the private test sets and to unseen categories; if any of these choices were fitted to the public set, the claimed margin will not generalize.

Editorial extensions

If this is right

If these results hold, the primary bottleneck in industrial anomaly detection moves from training algorithms to the quality of pre-trained visual features, since a training-free retriever beats trained detectors.
The small drop from 47.8% to 43.2% under mixed lighting, versus the large drops for trained methods like MSFlow (21.8% to 9.0%), would indicate that nearest-neighbor matching on robust features degrades gracefully under illumination shift.
The per-category memory bank of only 16 normal images implies that few-shot deployment on a new product line may need no gradient training, only a handful of defect-free images and a forward pass.
The near-zero baseline scores on categories such as Wallplugs (0.0%) imply that previously published methods essentially fail at those scenarios, making the reported 13.7% a substantial practical improvement even if far from perfect.

Reading between the lines

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

The reported margin probably over-attributes the improvement to training-free matching: two categories (Fabric, Walnuts) receive a fill-closed-regions post-processing that assumes anomalies are enclosed by normal-looking structure, and the F1 decision threshold is tuned on the public test set; a fair transfer test would freeze both.
A natural extension the authors do not explore is to test the same pipeline on other anomaly benchmarks (e.g., VisA, BTAD) without category-specific post-processing, which would isolate how much of the win comes from DINOv2 features versus the hand-built heuristics.
The PCA foreground-masking component suggests a cheap way to inject object-level priors into retrieval-based detectors; one could combine it with rough object proposals from a generic segmentation model to avoid per-category manual selection.
The paper itself documents failure cases (bubbles, specular highlights, missing-type anomalies) that bound the claim; a fair reader should view the quoted margins as category-dependent.

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.

Desk Editor's Note

read the letter

The short version: this paper claims a training-free DINOv2 memory-bank method beats every trained baseline on MVTec AD 2 by a huge margin in segmentation F1. The method is plausible and the writing is clear, but the evaluation has enough suspect choices that I would not take the headline numbers at face value yet.

What is actually new is modest: a few-shot greedy coreset memory bank built from DINOv2-L CLS tokens, features from layers 6/12/18/24, a PCA-based foreground mask for Vial and Wallplugs, and closed-region filling for Fabric and Walnuts. The combination is a natural extension of PatchCore and APRIL-GAN. The paper does not overclaim novelty and is honest about the failure cases. As a challenge technical report, it is fine.

The soft spots are real. In Section 2.2.2 the authors say they optimize the decision threshold on TESTpublic; that is test-set tuning, and it directly affects the reported F1 numbers. No ablations accompany the PCA mask or the category-specific post-processing, so it is hard to tell which components drive the gains, and the post-processing for Fabric/Walnuts looks like it was added after inspecting failure cases on the public test set. The comparison with baselines uses numbers taken from the MVTec AD 2 paper, and the evaluation protocol (binarization and thresholding) may not be identical. The gap is so large — 47.8 versus 21.8 mean F1 — that I want independent verification before believing it. No code is released, which makes that harder.

That said, the core idea is coherent and the method is simple enough to reproduce. The claims about being training-free are accurate in the sense of no gradient-based training. The paper is honestly scoped as a workshop report.

I would send this to a serious referee if it were submitted as a paper, but with the expectation that the evaluation is tightened: a proper validation split for threshold selection, ablations for the components, and ideally code release. For a challenge report, it is a useful data point. For a paper claiming consistent dominance over trained methods, it is not yet convincing.

Worth bringing to a reading group if you track anomaly detection benchmarks; I would cite it cautiously as a related result.

Referee Report

4 major / 5 minor

Summary. The paper proposes SuperAD, a fully training-free anomaly classification and segmentation method for the MVTec AD 2 challenge. The method uses a frozen DINOv2-ViT-L-14 backbone to extract features from four layers (6, 12, 18, 24) of a small memory bank of 16 normal reference images selected by greedy coreset sampling, computes per-layer nearest-neighbor anomaly maps, and averages them into a final segmentation map. Category-specific preprocessing is applied to Vial and Wallplugs (PCA-based foreground masking) and to Fabric and Walnuts (filling closed regions). The authors report segmentation F1 scores on TESTpublic, TESTpriv, and TESTpriv,mix and claim that the method consistently outperforms trained baselines such as EfficientAD, PatchCore, and MSFlow.

Significance. If the reported results are verified, the significance is considerable: a training-free memory-bank method built on DINOv2 features would outperform trained anomaly detectors on a challenging real-world benchmark, with practical advantages in deployment cost and generalization. The paper is clearly written, the method is simple and largely reproducible from the text, and the authors are explicit about the threshold-optimization procedure and about typical failure cases, which is a commendable disclosure. However, the central claim is currently supported only by a single evaluation with undisclosed threshold choices, no ablations, and baseline numbers taken from an external paper; the evidence is therefore not yet sufficient to establish the claimed level of superiority.

major comments (4)

[§2.2.2, Tables 1–2] The F1 decision threshold is explicitly optimized on TESTpublic. Since F1 is a threshold-dependent metric, the reported TESTpriv values (47.8% and 43.2%) may reflect information leakage from the public test set, and the large gap over MSFlow (21.8% and 9.0%) could be partly an artifact of threshold overfitting. Please report how the threshold was chosen (global or per-category), show the sensitivity of F1 to the threshold, or fix the threshold using a validation split from the training data; also provide error bars if any repeated runs were performed.
[Table 2, §3.2] The claim in Section 3.2 that SuperAD 'consistently outperforms previous methods' is contradicted by the paper's own Table 2: on Fruit Jelly for TESTpriv, MSFlow achieves 47.6% while SuperAD achieves 41.3%. This is a load-bearing discrepancy in the central claim. Please revise the claim to reflect average performance or explain the per-category comparison in detail.
[§3.1 vs §3.2] The reported numbers are internally inconsistent: Section 3.1 states that the official challenge server gives F1 scores of 47.18% and 42.51% on TESTpriv and TESTpriv,mix, while Table 2 lists 47.8% and 43.2% for the same sets. Please clarify which numbers are the official ones, which are locally recomputed, and why they differ; this ambiguity affects the validity of the headline comparison.
[§2.1.2, §2.2.1, Table 2] No ablation study isolates the contribution of the core components: the four-layer feature fusion, the greedy coreset selection, the memory bank size of 16, the PCA foreground masking applied only to Vial and Wallplugs, and the closed-region filling applied only to Fabric and Walnuts. The category-specific post-processing appears to have been selected based on observations on the public test set, and without an ablation or a held-out category analysis the reported gains cannot be attributed to the method rather than to the hand-picked post-processing rules.

minor comments (5)

[Title/Author block] The workshop name appears as 'V AND 3.0' with a spurious space; it should read 'VAND 3.0'.
[§2.2.1, Eq. (3)] In Equation (3), the left-hand side and the first branch of the right-hand side both use the symbol M_init, which is confusing; please use a different symbol for the corrected mask.
[§2.1.2] The name 'Patchcore' should be written as 'PatchCore' to match the reference list and standard usage.
[§2.2.2, §1.2] The metric 'AU-ROC 0.05' is used without a definition, and the notation later switches to 'AucPro0.05' and 'AU-PRO 0.05'; please define these metrics once and use a single consistent notation.
[Figure 1] The four 'NN Search' blocks in Figure 1 are not labeled with the layer indices (6, 12, 18, 24); adding these labels would make the architecture figure much easier to read.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SuperAD is a direct nearest-neighbor pipeline and no claimed result is defined by a fitted input.

full rationale

The paper's central claim is an empirical comparison on MVTec AD 2, not a derivation. The algorithm is fully specified: greedy coreset selection builds a 16-image memory bank from the category training set (Sec. 2.1.2), DINOv2-l/14 features from layers 6, 12, 18, 24 are matched by nearest-neighbor search, and the four similarity maps are averaged (Sec. 2.1.2-2.1.3). No equation in the paper reduces an output to an input by construction; the only fitted quantity is the F1 decision threshold, which is explicitly optimized on TESTpublic (Sec. 2.2.2). That is a threshold-selection protocol issue, not a circular derivation: the TESTpriv and TESTpriv,mix F1 numbers reported in Table 2 were obtained from the challenge server on separate held-out sets, so they are not forced by the TESTpublic threshold fit. The per-category post-processing choices (PCA foreground masking for Vial/Wallplugs, closed-region filling for Fabric/Walnuts) are heuristic and could reflect selection on observed results, but the paper does not define the evaluation result in terms of those choices; they are data-dependent preprocessing, not fitted parameters that determine the metric algebraically. The comparison baselines are external (PatchCore, RD, EfficientAD, MSFlow, SimpleNet, DSR), and the references to DINOv2, PatchCore, and DMAD are standard prior work, not a self-citation chain that supplies the load-bearing content. Therefore, despite legitimate concerns about threshold tuning and undisclosed ablations, no specific circular step can be exhibited, and the honest finding is no significant circularity with score 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The method rests on DINOv2's feature quality and several hand-set hyperparameters. The free parameters are mostly architectural choices (memory bank size, layers, resize) and the PCA threshold, plus the F1 threshold that is fit on the public test set. No new physical or conceptual entities are introduced. The axioms are standard domain assumptions for memory-bank anomaly detection.

free parameters (6)

memory_bank_size = 16
Number of reference images per category chosen by hand; no sensitivity analysis is provided.
feature_layers = [6, 12, 18, 24]
DINOv2 layers selected for multi-level feature extraction; no ablation justifies this choice.
resize_short_side = 672 (448 for Sheet Metal)
Input resolution for memory efficiency; chosen by hand and not varied.
pca_threshold_tau = 1.0
Threshold for binarizing the first PCA component to create the foreground mask; claimed to generalize without tuning but no cross-category evidence is given.
morphological_kernel_size = 3x3
Kernel used for dilation and erosion in mask post-processing; fixed without analysis.
f1_decision_threshold = optimized on TESTpublic
The threshold for binarizing the anomaly map to compute F1 is explicitly tuned on the TESTpublic set, making the reported F1 optimistic.

assumptions (5)

domain assumption DINOv2 features from layers 6/12/18/24 provide sufficient discriminative power to separate normal and anomalous regions via nearest neighbor matching.
Core premise of the method, stated in Sections 2.1.1-2.1.2; no analysis of feature separability is provided.
domain assumption Normal regions in a test image find a close neighbor in the 16-image memory bank, while anomalous regions do not.
Underlies the anomaly map construction in Section 2.1.2; this is the central assumption of memory-bank AD.
domain assumption The first PCA component of DINOv2 features separates foreground from background, and threshold tau=1.0 works across all categories.
Stated in Section 2.2.1; tested only on select categories and no sensitivity analysis is given.
domain assumption Greedy coreset selection on CLS tokens yields a representative memory bank of 16 images.
Section 2.1.2 assumes coreset diversity improves coverage of the normal manifold; no comparison to random selection is shown.
domain assumption F1 score on binarized images is the appropriate evaluation metric for MVTec AD 2.
Section 2.2.2 justifies F1 over AU-ROC for small defects, but the challenge may use additional or different metrics.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SuperAD: A Training-free Anomaly Classification and Segmentation Method for CVPR 2025 VAND 3.0 Workshop Challenge Track 1: Adapt & Detect." pith.science (2026). https://pith.science/paper/L5YN4HM3

@misc{pith2026250519750,
  author       = {Pith},
  title        = {Pith review of: SuperAD: A Training-free Anomaly Classification and Segmentation Method for CVPR 2025 VAND 3.0 Workshop Challenge Track 1: Adapt & Detect},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L5YN4HM3}},
  note         = {Machine review of arXiv:2505.19750}
}

read the original abstract

In this technical report, we present our solution to the CVPR 2025 Visual Anomaly and Novelty Detection (VAND) 3.0 Workshop Challenge Track 1: Adapt & Detect: Robust Anomaly Detection in Real-World Applications. In real-world industrial anomaly detection, it is crucial to accurately identify anomalies with physical complexity, such as transparent or reflective surfaces, occlusions, and low-contrast contaminations. The recently proposed MVTec AD 2 dataset significantly narrows the gap between publicly available benchmarks and anomalies found in real-world industrial environments. To address the challenges posed by this dataset--such as complex and varying lighting conditions and real anomalies with large scale differences--we propose a fully training-free anomaly detection and segmentation method based on feature extraction using the DINOv2 model named SuperAD. Our method carefully selects a small number of normal reference images and constructs a memory bank by leveraging the strong representational power of DINOv2. Anomalies are then segmented by performing nearest neighbor matching between test image features and the memory bank. Our method achieves competitive results on both test sets of the MVTec AD 2 dataset.

Figures

Figures reproduced from arXiv: 2505.19750 by the authors.

**Figure 2.** Examples of typical failure cases: (a) false positives on [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 8 canonical work pages

[1]

Efficien- tad: Accurate visual anomaly detection at millisecond-level latencies

Kilian Batzner, Lars Heckler, and Rebecca K ¨onig. Efficien- tad: Accurate visual anomaly detection at millisecond-level latencies. In Proceedings of the IEEE/CVF Winter Confer- ence on Applications of Computer Vision , pages 128–138,

work page
[2]

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. 2

arXiv 2023
[3]

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 Inter- national conference on pattern recognition, pages 475–489. Springer, 2021. 2

work page 2021
[4]

Anomaly detection via reverse distillation from one-class embedding

Hanqiu Deng and Xingyu Li. Anomaly detection via reverse distillation from one-class embedding. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9737–9746, 2022. 5

work page 2022
[5]

An image is worth 16x16 words: Trans- formers for image recognition at scale

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 2

arXiv 2010
[6]

The mvtec ad 2 dataset: Advanced scenarios for unsupervised anomaly detection

Lars Heckler-Kram, Jan-Hendrik Neudeck, Ulla Scheler, Re- becca K ¨onig, and Carsten Steger. The mvtec ad 2 dataset: Advanced scenarios for unsupervised anomaly detection. arXiv preprint arXiv:2503.21622, 2025. 1, 2

arXiv 2025
[7]

Dmad: Dual memory bank for real-world anomaly detection

Jianlong Hu, Xu Chen, Zhenye Gan, Jinlong Peng, Shengchuan Zhang, Jiangning Zhang, Yabiao Wang, Chengjie Wang, Liujuan Cao, and Rongrong Ji. Dmad: Dual memory bank for real-world anomaly detection. arXiv preprint arXiv:2403.12362, 2024. 2

arXiv 2024
[8]

Simplenet: A simple network for image anomaly detection and localization

Zhikang Liu, Yiming Zhou, Yuansheng Xu, and Zilei Wang. Simplenet: A simple network for image anomaly detection and localization. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 20402–20411, 2023. 5

work page 2023

Show all 14 references

[9]

Dinov2: Learning robust visual features without supervision

Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 2

2023 arXiv
[10]

Towards to- tal recall in industrial anomaly detection

Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Sch¨olkopf, Thomas Brox, and Peter Gehler. Towards to- tal recall in industrial anomaly detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14318–14328, 2022. 2, 5

2022
[11]

Revisiting reverse distillation for anomaly detection

Tran Dinh Tien, Anh Tuan Nguyen, Nguyen Hoang Tran, Ta Duc Huy, Soan Duong, Chanh D Tr Nguyen, and Steven QH Truong. Revisiting reverse distillation for anomaly detection. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 24511–2452...

2023
[12]

Wide residual net- works

Sergey Zagoruyko and Nikos Komodakis. Wide residual net- works. arXiv preprint arXiv:1605.07146, 2016. 2

2016 arXiv
[13]

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

Vitjan Zavrtanik, Matej Kristan, and Danijel Sko ˇcaj. Dsr– a dual subspace re-projection network for surface anomaly detection. In European conference on computer vision, pages 539–554. Springer, 2022. 5

2022
[14]

Msflow: Multiscale flow-based framework for unsupervised anomaly detection

Yixuan Zhou, Xing Xu, Jingkuan Song, Fumin Shen, and Heng Tao Shen. Msflow: Multiscale flow-based framework for unsupervised anomaly detection. IEEE Transactions on Neural Networks and Learning Systems, 2024. 2, 5

2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Efficien- tad: Accurate visual anomaly detection at millisecond-level latencies

Kilian Batzner, Lars Heckler, and Rebecca K ¨onig. Efficien- tad: Accurate visual anomaly detection at millisecond-level latencies. In Proceedings of the IEEE/CVF Winter Confer- ence on Applications of Computer Vision , pages 128–138,

work page

[2] [2]

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. 2

arXiv 2023

[3] [3]

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 Inter- national conference on pattern recognition, pages 475–489. Springer, 2021. 2

work page 2021

[4] [4]

Anomaly detection via reverse distillation from one-class embedding

Hanqiu Deng and Xingyu Li. Anomaly detection via reverse distillation from one-class embedding. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9737–9746, 2022. 5

work page 2022

[5] [5]

An image is worth 16x16 words: Trans- formers for image recognition at scale

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 2

arXiv 2010

[6] [6]

The mvtec ad 2 dataset: Advanced scenarios for unsupervised anomaly detection

Lars Heckler-Kram, Jan-Hendrik Neudeck, Ulla Scheler, Re- becca K ¨onig, and Carsten Steger. The mvtec ad 2 dataset: Advanced scenarios for unsupervised anomaly detection. arXiv preprint arXiv:2503.21622, 2025. 1, 2

arXiv 2025

[7] [7]

Dmad: Dual memory bank for real-world anomaly detection

Jianlong Hu, Xu Chen, Zhenye Gan, Jinlong Peng, Shengchuan Zhang, Jiangning Zhang, Yabiao Wang, Chengjie Wang, Liujuan Cao, and Rongrong Ji. Dmad: Dual memory bank for real-world anomaly detection. arXiv preprint arXiv:2403.12362, 2024. 2

arXiv 2024

[8] [8]

Simplenet: A simple network for image anomaly detection and localization

Zhikang Liu, Yiming Zhou, Yuansheng Xu, and Zilei Wang. Simplenet: A simple network for image anomaly detection and localization. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 20402–20411, 2023. 5

work page 2023