Pith. sign in

REVIEW 5 major objections 8 minor 56 references

Towards Zero-shot 3D Anomaly Localization

T0 review · 5 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read The paper claims 3D anomaly localization can transfer to unseen classes without target-class normal data, using pseudo-anomalies from a random CNN's attention on task-irrelevant point clouds.

desk verdict Defines a useful zero-shot 3D AD task and shows a plausible pipeline, but the headline SOTA comparison is confounded by asymmetric data access. read the letter →

arxiv 2412.04304 v1 pith:TV3VGFAP submitted 2024-12-05 cs.CV

classification cs.CV
keywords zero-shotanomalydetection3DlocalizationpointcloudpseudogenerationcontrastivelearningrandomCNNinductivebiasnormalcyclassifieradversarialperturbation
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

This paper identifies a new problem in industrial inspection: localizing anomalies in 3D point clouds of an object class for which no normal training data exist, because the training and testing classes do not overlap. It proposes 3DzAL, a patch-level contrastive learning framework that trains on normal data from one class and on pseudo anomalies synthesized from task-irrelevant classes, then tests on a disjoint class. The central move is using a randomly initialized, untrained CNN to highlight locations of interest in xyz point clouds; those points are attached to or removed from normal patches to create fake defects. Combined with a normalcy classifier and an adversarial perturbation of the classifier input, the framework claims to outperform the previous state of the art in both pixel-level and image-level zero-shot anomaly detection on the standard ten-class benchmark. If the claim holds, a factory could deploy off-the-shelf defect localization for new 3D parts without collecting or sharing normal samples of those parts.

What carries the argument

The load-bearing object is the pseudo-anomaly generation module. Given ordered xyz data from task-irrelevant classes, an untrained, randomly initialized CNN produces multi-scale activation maps; the highest-activation $\tau$ percent of points are treated as places of interest. Those points are attached to normal patches (adding-point anomalies) or a random patch has points removed (removing-point anomalies), in a 1:1 ratio. These pseudo anomalies feed a patch-level contrastive loss that pulls normal patch features together and pushes pseudo-anomaly features away, plus a representation-disentanglement loss that makes the learned 3D features complementary to FPFH features. A PointNet++ normalcy classifier is trained on the same normal-versus-pseudo-anomaly pairs with binary cross-entropy, and at test time its input is shifted against the gradient of its own softmax score before the classification probability is used as an anomaly score. The final score is a weighted sum of the memory-bank distance score and this perturbed classification score.

What would settle it

Run the same pipeline but replace the random-CNN point selection with the same number of uniformly random points; if pixel-level AUPRO does not drop noticeably, the claimed inductive bias is not what drives transfer.

Watch

Extended reading notes

Core claim

3DzAL establishes that the normal-versus-abnormal distinction can be learned class-agnostically. On the ten-class 3D industrial inspection benchmark, the paper systematically leaves one class out: it trains on the normal data of one class, uses the remaining eight as pseudo-anomaly sources, and tests on a ninth unseen class, covering 90 train/test pairs. The learned representation combines RGB features, handcrafted FPFH features, and PointNet++ features trained by contrastive learning against pseudo anomalies; a separate normalcy classifier scores patch abnormality, and the final score fuses a memory-bank distance with the classifier output on adversarially perturbed patch coordinates. The paper reports that 3DzAL beats the two strongest prior methods (BTF and 3DSR) on every category for pixel-level AUPRO and on most categories for image-level AUROC, with the largest per-category gains on localization. It also reports that the random, untrained CNN's attention on xyz data localizes points of interest better than an image-pretrained CNN for this task.

Load-bearing premise

The whole training signal rests on the assumption that pseudo-anomaly patches, built from points flagged by a randomly initialized CNN on task-irrelevant xyz data plus added or removed points, resemble the anomalies of the never-seen target class closely enough for the learned normal-versus-abnormal distinction to transfer.

Editorial extensions

If this is right

  • Anomaly localization can be done off-the-shelf for a new object class without collecting or sharing its normal data, which removes a practical barrier when target-class data are private or export-controlled.
  • The randomly initialized CNN's attention on xyz data is a free prior for 3D geometric interest points; it can be used anywhere a cheap saliency map over point clouds is needed.
  • The method's training signal depends only on generic geometric edits to normal patches, so the same recipe should transfer to other 3D anomaly benchmarks beyond the ten-class set studied here.
  • Because the learned features are complementary to FPFH, the framework does not need a 3D-pretrained model; the only pretrained component is the standard RGB backbone.
  • Fusing the distance score with the perturbed classifier score improves both localization and detection, suggesting that the two scores capture partially independent evidence.

Reading between the lines

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

  • Because the pseudo anomalies are class-agnostic geometric edits (adding and removing point clusters), the same generation recipe could be recycled for other 3D inspection tasks, such as detecting deformation in non-rigid objects, where the notion of 'normal shape' is class-specific.
  • A directly testable extension is to increase the diversity of task-irrelevant classes or add synthetic CAD data; if the transfer gains grow with pseudo-anomaly diversity, the core mechanism is confirmed as cross-class geometric contrast.
  • The adversarial perturbation on the classifier input is a test-time sharpening step; one could check whether it also improves the classical same-class setting, which would separate its value from zero-shot transfer specifically.
  • The fixed score weights $w_d$ and $w_c$ are set once in the paper; a probing experiment could check whether per-target-class tuning of these weights changes the ranking across categories, since the current averages may mask category-dependent sensitivity.
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 / 8 minor

Summary. The paper introduces the task of zero-shot 3D anomaly detection and localization, where the training class and test class do not overlap, and proposes 3DzAL. 3DzAL uses a randomly initialized, untrained CNN on ordered xyz point clouds to generate pseudo-anomaly patches from eight task-irrelevant classes, trains a PointNet++ feature extractor with contrastive and disentanglement losses, trains a normalcy classifier, and combines a PatchCore-style distance score with an adversarially perturbed classification score. Experiments on MVTec 3D-AD follow a leave-one-out protocol over 10 classes (90 train/test pairs) and report pixel-level AUPRO and image-level AUROC against BTF and 3DSR adapted to the zero-shot setting, with ablations over components, pseudo-anomaly types, and CNN initialization.

Significance. The proposed task is practically motivated and the paper is one of the first to address it. If the reported comparisons were controlled, the framework would be a solid zero-shot baseline. The observation that a random CNN on xyz input produces activation maps that highlight anomalous-looking regions (Fig. 3) is interesting and potentially reusable beyond anomaly detection. Strengths include the exhaustive 90-experiment leave-one-out evaluation, explicit ablations (Tables 3-6), and comparison of model/memory-bank sizes. However, the main empirical claim is currently confounded by unequal access to auxiliary data and by unreported final-score weights, so the significance cannot be fully assessed until those issues are addressed.

major comments (5)
  1. [Section 4, Experimental setting; Tables 1 and 2] The leave-one-out protocol gives 3DzAL access to eight task-irrelevant classes for pseudo-anomaly generation in addition to the single training class, while the adapted BTF and 3DSR baselines are trained or built from that single training class only. The reported AUPRO/AUROC gains are therefore not attributable to the proposed cross-class transfer mechanism; they may instead reflect a 9-class versus 1-class data budget. Please rerun BTF and 3DSR with the same nine non-test classes (e.g., a BTF memory bank built from all nine classes and 3DSR trained on all nine) and compare, or ablate 3DzAL with reduced auxiliary data, so that the comparison controls for data access.
  2. [Section 3.4, Eqs. (11)-(12); Implementation details] The final anomaly score uses weights wd and wc, but their values are never reported and no sensitivity analysis is provided. Since the reported AUPRO is computed from this weighted sum, the unspecified weights make the main tables non-reproducible and leave open the possibility of per-category tuning. Please report the values used for wd and wc and demonstrate that the conclusions are stable over a range of these weights.
  3. [Section 3.2 and Table 5] The 'inductive bias' of the random CNN is a core component of the method, but the paper does not compare pseudo-anomaly generation against a trivial baseline of randomly selected patches. Table 5 only compares random versus ImageNet-initialized CNN weights; without a random-patch control, the claim that the random CNN localizes anomaly-relevant regions (contribution 3) is not quantitatively supported. Please add such a control and, if possible, quantify the agreement between the activation mask A and ground-truth anomaly locations on task-irrelevant data.
  4. [Section 2 and Section 4, Baselines] The paper explicitly states that it 'intentionally do[es] not make use of any existing zero-shot AD work' and therefore omits 2D zero-shot anomaly-detection baselines. Because 3DzAL has an RGB branch and the dataset provides RGB images, comparisons with simple zero-shot 2D methods (e.g., ImageNet-feature distance or CLIP-based scores on RGB) are feasible and necessary to support the statement that the method outperforms 'state-of-the-art anomaly detection and localization' in a zero-shot setting. Please add at least one such baseline or justify why RGB-only zero-shot methods are outside the comparison scope.
  5. [Tables 1-3 and Implementation details] No error bars or statistical significance are reported for the 90 individual experiments, and several key hyperparameters (tau, negative-to-positive patch ratio, T, eta, b, patch size, epochs) are fixed without sensitivity analysis. Given the modest per-component gains in Table 3 (e.g., the contrastive loss adds no more than about 0.2 AUPRO in the reported settings), reporting variance over at least three random seeds and a sensitivity sweep for the most critical hyperparameters (tau, eta, b, and wd/wc) would materially strengthen the claim that the improvements are robust.
minor comments (8)
  1. [Abstract and author list] There are typographical artifacts such as 'W e' and 'Y un Fu' in the author block; these should be corrected.
  2. [Section 3.2] The phrase 'reciprocal second, third, and fourth layer output' should be 'respective second, third, and fourth layer outputs', and 'comparitively' is a typo for 'comparatively'.
  3. [Section 3.4] The term 'KNN Gaussian Blurring' is unclear; if this refers to a kernel/Gaussian blur used to smooth the score map, it should be stated explicitly and distinguished from the nearest-neighbor parameter b.
  4. [References] Reference [22] is listed as '3D-ADS. 2022.' with no title or venue; it should be completed.
  5. [Section 4 and Table 4] The text says the ablations are run on bagel, dowel, and foam, but Table 4 reports bagel, potato, and rope; please clarify which training classes were used for each ablation table.
  6. [Section 4, Tables 1 and 2] The row/column structure reports means over the nine test classes per training class, but the captions say 'in all categories'/'in most categories'; please specify that these statements refer to the per-training-class means, since the individual test-class entries are not compared separately in the table.
  7. [Section 3.4, Eq. (8)] The definition of bp(xtest|w) as max{p(xtest|w), 1-p(xtest|w)} makes the sign of the perturbation step hard to follow; please clarify the predicted class and state explicitly why adding -eta times the gradient lowers the confidence of the predicted class.
  8. [Implementation details] Please state how the 'ordered' 3D point cloud is obtained from the MVTec 3D-AD data (e.g., whether it is the native organized point cloud or the result of a projection/unprojection step), as this is needed for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: 3DzAL's zero-shot transfer claim is evaluated on external MVTec 3D-AD data, and no prediction reduces to a fitted value or to a self-citation by construction.

full rationale

The paper's derivation chain is not circular. Pseudo anomalies are synthesized from task-irrelevant classes using a frozen, randomly initialized CNN; no parameter is fitted to the target class, and target anomalies are unseen during training. The contrastive loss, normalcy classifier, and adversarial perturbation are all trained only on normal training patches plus synthetic pseudo anomalies from non-test classes, and the final comparison is against BTF and 3DSR on the external MVTec 3D-AD benchmark. The self-citations (Aich et al. [1], which shares author Kuan-Chuan Peng, and the authors' prior works) provide the 2D inductive-bias and perturbation techniques but are not used to justify the zero-shot 3D result; that result rests on the reported experiments. A caveat is that the experimental protocol gives 3DzAL access to eight additional classes for pseudo-anomaly synthesis while the baselines see only one training class, which is an evaluation-fairness concern rather than a circularity. The paper's own limitation note (low AUPRO for foam-to-cookie) further confirms the result is not forced by construction.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The method is a hand-tuned pipeline: every component score depends on manually chosen hyperparameters, and the final combination weights are unreported. The core assumptions about random-CNN attention and cross-class transfer are not independently verified.

free parameters (10)
  • wd and wc (final anomaly score weights)
    Introduced in Eqs. 11-12 but never assigned values in Section 4; the reported anomaly scores depend on them.
  • tau (pseudo-anomaly point selection percentage) = 0.1%
    Section 4 Implementation details; chosen by hand, no sensitivity analysis.
  • negative-to-positive patch ratio = 16:1
    Section 4; manually set for contrastive learning and normalcy classifier.
  • temperature T = 0.07
    Section 4; standard contrastive temperature but a free choice.
  • loss weights wcon and wrd = wcon=1, wrd=100
    Section 4; chosen to make loss ranges comparable.
  • perturbation magnitude eta = 0.1
    Section 4; no sensitivity analysis.
  • nearest neighbors b in score reweighting = 3
    Section 4; manually chosen.
  • adding/removing pseudo anomaly ratio = 1:1
    Section 4; manually chosen.
  • patch size and input resolution = 8x8 patches, 224x224 ordered data
    Section 4; architecture choices that affect feature granularity.
  • training epochs = 5
    Section 4; last-epoch model used, no early stopping analysis.
assumptions (6)
  • ad hoc to paper A randomly initialized, untrained CNN applied to ordered xyz point clouds produces activation maps that localize anomaly-relevant regions.
    Used as the basis for pseudo-anomaly generation in Sec. 3.2; supported only by Fig. 3 visualizations and a three-class ablation in Tab. 5, not by an independent quantitative evaluation of mask quality.
  • domain assumption Pseudo anomalies synthesized from task-irrelevant classes transfer to the unseen target class.
    The entire zero-shot design (Secs. 3.2 and 3.3) assumes the normal versus pseudo-abnormal distinction learned on eight task-irrelevant classes is indicative of test anomalies in the held-out class.
  • domain assumption Concatenating FPFH, ImageNet RGB, and learned PointNet++ features gives complementary representations useful for anomaly scoring.
    The disentanglement loss in Eq. 2 forces complementarity to FPFH, but complementarity to the RGB branch is assumed, not verified.
  • domain assumption PatchCore-style memory bank and coreset distance scoring remain valid when training and testing classes differ.
    Section 3.4 directly imports PatchCore's nearest-neighbor and reweighting scheme without justification for cross-class distribution shift.
  • domain assumption MVTec 3D-AD with leave-one-class-out simulates the practical zero-shot deployment setting.
    Section 4 uses 90 train/test class splits on one benchmark; generalizability to other 3D data or anomaly types is not shown.
  • ad hoc to paper Adversarial input perturbation with a fixed eta improves the classification-based anomaly score.
    Eq. 8 and the score combination in Eq. 11 rely on this heuristic; its contribution is shown only by the ablation in Tab. 3 for three training classes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Zero-shot 3D Anomaly Localization." pith.science (2026). https://pith.science/paper/TV3VGFAP

@misc{pith2026241204304,
  author       = {Pith},
  title        = {Pith review of: Towards Zero-shot 3D Anomaly Localization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TV3VGFAP}},
  note         = {Machine review of arXiv:2412.04304}
}
read the original abstract

3D anomaly detection and localization is of great significance for industrial inspection. Prior 3D anomaly detection and localization methods focus on the setting that the testing data share the same category as the training data which is normal. However, in real-world applications, the normal training data for the target 3D objects can be unavailable due to issues like data privacy or export control regulation. To tackle these challenges, we identify a new task -- zero-shot 3D anomaly detection and localization, where the training and testing classes do not overlap. To this end, we design 3DzAL, a novel patch-level contrastive learning framework based on pseudo anomalies generated using the inductive bias from task-irrelevant 3D xyz data to learn more representative feature representations. Furthermore, we train a normalcy classifier network to classify the normal patches and pseudo anomalies and utilize the classification result jointly with feature distance to design anomaly scores. Instead of directly using the patch point clouds, we introduce adversarial perturbations to the input patch xyz data before feeding into the 3D normalcy classifier for the classification-based anomaly score. We show that 3DzAL outperforms the state-of-the-art anomaly detection and localization performance.

Figures

Figures reproduced from arXiv: 2412.04304 by the authors.

Figure 1
Figure 1. Problem overview. Current 3D anomaly detection and localization works entail training on the normal data of one class and testing on the normal and abnormal data of the same class. We extend such setting by testing on other classes without the corresponding normal training data. This zero-shot setting is practical when such data are unavailable (e.g., due to data privacy, export control laws, etc.). GT denotes groun… view at source ↗
Figure 2
Figure 2. Framework overview. Our proposed 3DzAL framework mainly adopts three branches to extract features given both 2D and 3D data of an object. The RGB branch extracts feature from 2D image data of the object using ResNet pre-trained on ImageNet. The FPFH branch extracts handcrafted FPFH features from 3D point cloud data. The point cloud branch employs a learnable network (PointNet++) to extract features. The network is t… view at source ↗
Figure 3
Figure 3. Inductive bias of random networks. We feed the xyz data of abnormal examples as the input of a randomly initialized and untrained ResNet-50, and visualize the attention maps. These maps show that the random network has the inductive bias of covering the locations of interest, including the locations shown in the ground truth. the output sizes of the reciprocal second, third, and fourth layers are 14×14, 28×28, and 5… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Pseudo anomaly generation. Overview of our proposed patch-level 3D pseudo anomaly sample generation process for both “adding” and “removing” type anomalies. To obtain the object-level anomaly score Sdist, we impose an additional weight in the following form: Sdist(Xtes…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 48 canonical work pages

  1. [1]

    Cross-domain video anomaly detection without target domain adaptation

    Abhishek Aich, Kuan-Chuan Peng, and Amit K Roy-Chowdhury. Cross-domain video anomaly detection without target domain adaptation. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V ision, pages 2579–2591, 2023. 3, 4

  2. [2]

    MVTec AD–a comprehensive real-world dataset for un- supervised anomaly detection

    Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. MVTec AD–a comprehensive real-world dataset for un- supervised anomaly detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9592–9600, 2019. 4

  3. [3]

    The MVTec 3D-AD dataset for unsupervised 3D anomaly detection and localization

    Paul Bergmann., Xin Jin., David Sattlegger., and Carsten Steger. The MVTec 3D-AD dataset for unsupervised 3D anomaly detection and localization. InProceedings of the 17th International Joint Conference on Computer V ision, Imaging and Computer Graphics Theory and Applications - V olume 5: VISAPP ,, pages 202–213. INSTICC, SciTePress, 2022. 1, 2, 4, 7

  4. [4]

    Anomaly detection in 3D point clouds using deep geometric descriptors

    Paul Bergmann and David Sattlegger. Anomaly detection in 3D point clouds using deep geometric descriptors. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V ision, pages 2613–2623, 2023. 2

  5. [5]

    Col- laborative discrepancy optimization for reliable image anomaly localization

    Y unkang Cao, Xiaohao Xu, Zhaoge Liu, and Weiming Shen. Col- laborative discrepancy optimization for reliable image anomaly localization. IEEE Transactions on Industrial Informatics, 2023. 1, 2

  6. [6]

    Complementary pseudo multimodal feature for point cloud anomaly detection

    Y unkang Cao, Xiaohao Xu, and Weiming Shen. Complementary pseudo multimodal feature for point cloud anomaly detection. arXiv preprint arXiv:2303.13194, 2023. 1

  7. [7]

    A random CNN sees objects: One inductive bias of cnn and its applications

    Y un-Hao Cao and Jianxin Wu. A random CNN sees objects: One inductive bias of cnn and its applications. InProceedings Of The AAAI Conference On Artificial Intelligence, volume 36, pages 194–202, 2022. 4

  8. [8]

    Transductive zero-shot learning for 3d point cloud classification

    Ali Cheraghian, Shafin Rahman, Dylan Campbell, and Lars Petersson. Transductive zero-shot learning for 3d point cloud classification. InProceedings of the IEEE/CVF winter conference on applications of computer vision, pages 923–933, 2020. 1

Show all 56 references
  1. [9]

    Zero-shot learning on 3d point cloud objects and beyond.International Journal of Computer V ision, 130(10):2364–2384, 2022

    Ali Cheraghian, Shafin Rahman, Townim F Chowdhury, Dylan Campbell, and Lars Petersson. Zero-shot learning on 3d point cloud objects and beyond.International Journal of Computer V ision, 130(10):2364–2384, 2022. 1

  2. [10]

    Shape-guided dual-memory learning for 3d anomaly detection

    Y u-Min Chu, Chieh Liu, Ting-I Hsieh, Hwann-Tzong Chen, and Tyng-Luh Liu. Shape-guided dual-memory learning for 3d anomaly detection. In International Conference on Machine Learning, pages 6185–6194. PMLR, 2023. 2

  3. [11]

    Sub-image anomaly detec- tion with deep pyramid correspondences

    Niv Cohen and Y edid Hoshen. Sub-image anomaly detec- tion with deep pyramid correspondences. arXiv preprint arXiv:2005.02357, 2020. 3

  4. [12]

    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. InP attern Recognition. ICPR International W orkshops and Challenges: V irtual Event, January 10–15, 2021, Proceeding...

  5. [13]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 7

  6. [14]

    Catching both gray and black swans: Open-set supervised anomaly detec- tion

    Choubo Ding, Guansong Pang, and Chunhua Shen. Catching both gray and black swans: Open-set supervised anomaly detec- tion. In Proceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 7388–7398, 2022. 3

  7. [15]

    Zero-shot out-of-distribution detection based on the pre-trained model CLIP

    Sepideh Esmaeilpour, Bing Liu, Eric Robertson, and Lei Shu. Zero-shot out-of-distribution detection based on the pre-trained model CLIP. InProceedings of the AAAI conference on artificial intelligence, volume 36, pages 6568–6576, 2022. 3

  8. [16]

    A graph-based one-shot learning method for point cloud recognition

    Zhaoxin Fan, Hongyan Liu, Jun He, Qi Sun, and Xiaoyong Du. A graph-based one-shot learning method for point cloud recognition. In Computer Graphics F orum, volume 39, pages 313–323. Wiley Online Library, 2020. 1

  9. [17]

    Mpdnet: A 3d missing part detection network based on point cloud segmentation

    Zhaoxin Fan, Hongyan Liu, Jun He, Min Zhang, and Xiaoyong Du. Mpdnet: A 3d missing part detection network based on point cloud segmentation. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1810–1814. IEEE, 2021. 1

  10. [18]

    Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit

    Richard HR Hahnloser, Rahul Sarpeshkar, Misha A Mahowald, Rodney J Douglas, and H Sebastian Seung. Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit. nature, 405(6789):947–951, 2000. 4

  11. [19]

    Masked autoencoders are scalable vision learn- ers

    Kaiming He, Xinlei Chen, Saining Xie, Y anghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learn- ers. In Proceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 16000–16009, 2022. 3

  12. [20]

    Delv- ing deep into rectifiers: Surpassing human-level performance on imagenet classification

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delv- ing deep into rectifiers: Surpassing human-level performance on imagenet classification. InProceedings of the IEEE international conference on computer vision, pages 1026–1034, 2015. 4

  13. [21]

    Long-tailed anomaly detection with learnable class names

    Chih-Hui Ho, Kuan-Chuan Peng, and Nuno V asconcelos. Long-tailed anomaly detection with learnable class names. In Proceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 12435–12446, 2024. 3

  14. [22]

    Eliahu Horwitz and Y edid Hoshen. 3D-ADS. 2022. 7

  15. [23]

    Back to the feature: Classical 3d features are (almost) all you need for 3d anomaly detection

    Eliahu Horwitz and Y edid Hoshen. Back to the feature: Classical 3d features are (almost) all you need for 3d anomaly detection. In Proceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition (CVPR) W orkshops, pages 2967–2976, June 2023. 1, 2, 4, 7

  16. [24]

    Registration based few-shot anomaly detection

    Chaoqin Huang, Haoyan Guan, Aofan Jiang, Y a Zhang, Michael Spratling, and Y an-Feng Wang. Registration based few-shot anomaly detection. In Computer V ision–ECCV 2022: 17th European Conference, T el Aviv , Israel, October 23–27, 2022, Proceedings, P art XXIV, pages 303–319. S...

  17. [25]

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

    Jongheon Jeong, Y ang Zou, Taewan Kim, Dongqing Zhang, Avinash Ravichandran, and Onkar Dabeer. Winclip: Zero-/few- shot anomaly classification and segmentation. InProceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 19606–19616, 2023. 3

  18. [26]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014. 7

  19. [27]

    V andermeulen, Billy Joe Franks, Marius Kloft, and Klaus Robert Muller

    Philipp Liznerski, Lukas Ruff, Robert A. V andermeulen, Billy Joe Franks, Marius Kloft, and Klaus Robert Muller. Explainable deep one-class classification. In International Conference on Learning Representations, 2021. 3

  20. [28]

    V andermeulen, Billy Joe Franks, Klaus Robert Muller, and Marius Kloft

    Philipp Liznerski, Lukas Ruff, Robert A. V andermeulen, Billy Joe Franks, Klaus Robert Muller, and Marius Kloft. Exposing outlier exposure: What can be learned from few, one, and zero outlier images. Transactions on Machine Learning Research, 2022. 3

  21. [29]

    Few-shot scene-adaptive anomaly detection

    Yiwei Lu, Frank Y u, Mahesh Kumar Krishna Reddy, and Y ang Wang. Few-shot scene-adaptive anomaly detection. InComputer V ision–ECCV 2020: 16th European Conference, Glasgow , UK, August 23–28, 2020, Proceedings, P art V 16, pages 125–141. Springer, 2020. 3

  22. [30]

    Toward unsupervised 3D point cloud anomaly detection using variational autoencoder

    Mana Masuda, Ryo Hachiuma, Ryo Fujii, Hideo Saito, and Y usuke Sekikawa. Toward unsupervised 3D point cloud anomaly detection using variational autoencoder. In2021 IEEE International Conference on Image Processing (ICIP), pages 3118–3122. IEEE, 2021. 1

  23. [31]

    Explainable deep few-shot anomaly detection with deviation networks.arXiv preprint arXiv:2108.00462, 2021

    Guansong Pang, Choubo Ding, Chunhua Shen, and Anton van den Hengel. Explainable deep few-shot anomaly detection with deviation networks.arXiv preprint arXiv:2108.00462, 2021. 3

  24. [32]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems, 30, 2017. 5, 7

  25. [33]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InInternational con- ference on machine learning, p...

  26. [34]

    Towards total recall in industrial anomaly detection

    Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Sch¨olkopf, Thomas Brox, and Peter Gehler. Towards total recall in industrial anomaly detection. InProceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 14318–14328, 2022. 2, 3, 4, 5

  27. [35]

    Asymmetric student-teacher networks for industrial anomaly detection

    Marco Rudolph, Tom Wehrbein, Bodo Rosenhahn, and Bastian Wandt. Asymmetric student-teacher networks for industrial anomaly detection. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V ision, pages 2592–2602, 2023. 1, 2

  28. [36]

    Fast point feature histograms (fpfh) for 3d registration

    Radu Bogdan Rusu, Nico Blodow, and Michael Beetz. Fast point feature histograms (fpfh) for 3d registration. In2009 IEEE international conference on robotics and automation, pages 3212–3217. IEEE, 2009. 2, 4

  29. [37]

    Maeday: Mae for few- and zero-shot anomaly-detection.Computer V ision and Image Understanding, 241:103958, 2024

    Eli Schwartz, Assaf Arbelle, Leonid Karlinsky, Sivan Harary, Florian Scheidegger, Sivan Doveh, and Raja Giryes. Maeday: Mae for few- and zero-shot anomaly-detection.Computer V ision and Image Understanding, 241:103958, 2024. 3

  30. [38]

    Active learning for convolu- tional neural networks: A core-set approach

    Ozan Sener and Silvio Savarese. Active learning for convolu- tional neural networks: A core-set approach. InInternational Conference on Learning Representations, 2018. 5

  31. [39]

    A hierarchi- cal transformation-discriminating generative model for few shot anomaly detection

    Shelly Sheynin, Sagie Benaim, and Lior Wolf. A hierarchi- cal transformation-discriminating generative model for few shot anomaly detection. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer V ision, pages 8495–8504, 2021. 3

  32. [40]

    Small-gan: Speeding up gan training using core-sets

    Samarth Sinha, Han Zhang, Anirudh Goyal, Y oshua Bengio, Hugo Larochelle, and Augustus Odena. Small-gan: Speeding up gan training using core-sets. InInternational Conference on Machine Learning, pages 9005–9015. PMLR, 2020. 5

  33. [41]

    Attention guided anomaly localization in images

    Shashanka V enkataramanan, Kuan-Chuan Peng, Rajat Vikram Singh, and Abhijit Mahalanobis. Attention guided anomaly localization in images. InEuropean Conference on Computer V ision, pages 485–503. Springer, 2020. 3

  34. [42]

    Position encoding enhanced feature mapping for image anomaly detection

    Qian Wan, Y unkang Cao, Liang Gao, Weiming Shen, and Xinyu Li. Position encoding enhanced feature mapping for image anomaly detection. In2022 IEEE 18th International Conference on Automation Science and Engineering (CASE), pages 876–881. IEEE, 2022. 1, 2

  35. [43]

    To- wards explainable visual anomaly detection

    Yizhou Wang, Dongliang Guo, Sheng Li, and Y un Fu. To- wards explainable visual anomaly detection. arXiv preprint arXiv:2302.06670, 2023. 1

  36. [44]

    Multimodal industrial anomaly detection via hybrid fusion

    Y ue Wang, Jinlong Peng, Jiangning Zhang, Ran Yi, Y abiao Wang, and Chengjie Wang. Multimodal industrial anomaly detection via hybrid fusion. InProceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition, pages 8032–8041, 2023. 2

  37. [45]

    Making reconstruction-based method great again for video anomaly detection

    Yizhou Wang, Can Qin, Y ue Bai, Yi Xu, Xu Ma, and Y un Fu. Making reconstruction-based method great again for video anomaly detection. In2022 IEEE International Conference on Data Mining (ICDM), pages 1215–1220. IEEE, 2022. 1

  38. [46]

    Self-supervision meets adversarial perturbation: A novel framework for anomaly detection

    Yizhou Wang, Can Qin, Rongzhe Wei, Yi Xu, Y ue Bai, and Y un Fu. Self-supervision meets adversarial perturbation: A novel framework for anomaly detection. InProceedings of the 31st ACM International Conference on Information & Knowledge Management, pages 4555–4559, 2022. 1

  39. [47]

    Sla22p: Self-supervised anomaly detection with adversarial perturbation

    Yizhou Wang, Can Qin, Rongzhe Wei, Yi Xu, Y ue Bai, and Y un Fu. Sla22p: Self-supervised anomaly detection with adversarial perturbation. IEEE Transactions on Knowledge and Data Engineering, 36(12):9282–9293, 2024. 1

  40. [48]

    Learning unsupervised metaformer for anomaly detection

    Jhih-Ciang Wu, Ding-Jie Chen, Chiou-Shann Fuh, and Tyng-Luh Liu. Learning unsupervised metaformer for anomaly detection. In Proceedings of the IEEE/CVF International Conference on Computer V ision, pages 4369–4378, 2021. 3

  41. [49]

    IM-IAD: Industrial image anomaly detection benchmark in manufacturing

    Guoyang Xie, Jinbao Wang, Jiaqi Liu, Jiayi Lyu, Y ong Liu, Chengjie Wang, Feng Zheng, and Y aochu Jin. IM-IAD: Industrial image anomaly detection benchmark in manufacturing. arXiv preprint arXiv:2301.13359, 2023. 1

  42. [50]

    Bdd100k: A diverse driving dataset for heterogeneous multitask learning

    Fisher Y u, Haofeng Chen, Xin Wang, Wenqi Xian, Yingying Chen, Fangchen Liu, V ashisht Madhavan, and Trevor Darrell. Bdd100k: A diverse driving dataset for heterogeneous multitask learning. In Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition,...

  43. [51]

    Wide residual networks

    Sergey Zagoruyko and Nikos Komodakis. Wide residual networks. InBMVC, 2016. 7

  44. [52]

    Vitjan Zavrtanik, Matej Kristan, and Danijel Skoˇcaj. 3DSR

  45. [53]

    Cheating depth: Enhancing 3d surface anomaly detection via depth simulation

    Vitjan Zavrtanik, Matej Kristan, and Danijel Skoˇcaj. Cheating depth: Enhancing 3d surface anomaly detection via depth simulation. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V ision, pages 2164–2172, 2024. 2, 7

  46. [54]

    Benchmarking unsupervised anomaly detection and localization

    Y e Zheng, Xiang Wang, Y u Qi, Wei Li, and Liwei Wu. Benchmarking unsupervised anomaly detection and localization. arXiv preprint arXiv:2205.14852, 2022. 1

  47. [55]

    Real-time kd-tree construction on graphics hardware.ACM Transactions on Graphics (TOG), 27(5):1–11, 2008

    Kun Zhou, Qiming Hou, Rui Wang, and Baining Guo. Real-time kd-tree construction on graphics hardware.ACM Transactions on Graphics (TOG), 27(5):1–11, 2008. 4

  48. [56]

    Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection.arXiv preprint arXiv:2310.18961,

    Qihang Zhou, Guansong Pang, Y u Tian, Shibo He, and Jiming Chen. Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection.arXiv preprint arXiv:2310.18961,

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.