Pith. sign in

REVIEW 3 major objections 5 minor 32 references

MVFM-3DAD: Multi-view Flow Matching for 3D Anomaly Detection via Density Proxy Estimation

T0 review · 3 major / 5 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 Unsupervised 3D anomaly detection can be solved by learning how normal multi-view features flow toward a Gaussian and scoring test features by their terminal energy.

desk verdict A solid, well-engineered 3D anomaly detection paper that earns its Real3D-AD gains and deserves peer review, but its small MVTec margins and the unproved OOD transport assumption behind Eq. (13) need attention. read the letter →

arxiv 2608.12148 v1 pith:TNRQLVWR submitted 2026-08-12 cs.GR

Liangwei Li , Lin Liu , Jing Zhang , Xiaohui Du , Ruqian Hao , Xinwei Li , Hanzhe Liang , Juanxiu Liu This is my paper · ORCID
classification cs.GR
keywords 3Danomalydetectionflowmatchingdensityproxyestimationmulti-viewprojectionpointcloudlocalizationunsupervisedlearningterminalGaussianenergy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that 3D anomaly detection and localization can be done without memory banks or reconstruction: normality is modeled as a learned transport from normal multi-view features to a Gaussian reference, and anomalies are features whose transported endpoint falls in a low-probability region of that Gaussian. To do this, MVFM-3DAD projects each point cloud into multiple rendered views, trains a flow-matching velocity field on normal view features only, and maps the resulting per-view density-proxy scores back to 3D points using the pixel-to-point correspondences recorded during projection. A sympathetic reader would care because, if correct, the approach sidesteps two known failure modes—finite memory coverage and identity-shortcut reconstruction—while using a frozen pretrained visual encoder. The paper reports state-of-the-art results on two standard 3D anomaly benchmarks, with mean O-AUROC/P-AUROC of 90.8%/96.0% on Real3D-AD and 95.9%/95.3% on MVTec3D-AD.

What carries the argument

The load-bearing mechanism is the pair of a Bidirectional Geometric Projector (BGP) and a Flow-guided Density Proxy Estimator (FDPE). BGP is a deterministic interface: forward, it rotates each point cloud around its center for a set of viewpoints and renders one image per view while recording the projected pixel coordinate of every 3D point; backward, it bilinearly upsamples the per-view anomaly evidence and samples it at those stored coordinates, averaging over views and channels to give each point a score. FDPE is the density model. Given a normalized view feature $X$, it samples a Gaussian endpoint $\epsilon \sim \mathcal{N}(0,I)$, builds the linear transport path $X_t = (1-t)X + t\epsilon$, and trains a time-conditioned velocity field $g_\psi$ to match the target direction $\epsilon - X$ on normal samples only. At inference it solves the learned ODE with Euler steps and scores the terminal state by $1-\exp(-Z_{\text{Node}}^2/2)$. This machinery converts an irregular point cloud into a regular feature grid where a standard flow model applies, and then converts the flow evidence back into point-level scores without any learnable decoder.

What would settle it

Take a normal point cloud, add synthetic surface defects of increasing size, and plot the per-point score $1-\exp(-Z_{\text{Node}}^2/2)$ for the affected view features. If the score does not rise monotonically with defect size while staying at the normal baseline for undamaged regions on a held-out class, the terminal-energy criterion is not the monotone anomaly score the paper assumes.

Watch

Extended reading notes

Core claim

The central claim is that unsupervised 3D anomaly detection can be reframed as density-proxy estimation in a structured multi-view feature space. A point cloud is rendered from several viewpoints; a frozen vision encoder turns each view into a feature map; and a flow-matching module learns to transport those normal features along linear paths to a standard Gaussian reference $\epsilon \sim \mathcal{N}(0,I)$. At test time the same learned velocity field is integrated from the test feature, and the terminal state is evaluated by the Gaussian energy $1-\exp(-Z_{\text{Node}}^2/2)$: high terminal energy means the feature does not follow the normal distribution. Because the projector records where each 3D point lands in every view, those per-view energies are sampled back at the stored coordinates and averaged into a point-level anomaly map, whose maximum gives the object score. The paper asserts that this density-proxy criterion, without input reconstruction or explicit Jacobian evaluation, outperforms memory-bank and reconstruction baselines on Real3D-AD and MVTec3D-AD.

Load-bearing premise

The whole scoring relies on the unproven premise that when an anomalous feature is pushed along the velocity field learned only from normal features, it lands in a higher-energy, rarer part of the terminal Gaussian than a normal feature would.

Editorial extensions

If this is right

  • Because the anomaly score never reconstructs the input, the identity-shortcut failure that lets high-capacity reconstruction networks reproduce anomalous regions is bypassed by construction.
  • Point-level localization is tied to the exact projection geometry: evidence from any view that sees a defect contributes to the same 3D point, so the method localizes defects even when they are visible in only one of the rendered views.
  • At inference, scoring costs only a small number of Euler steps of a learned velocity field and no Jacobian or divergence evaluation, so the mechanism is inexpensive for high-dimensional features.
  • On the two evaluated benchmarks, the reported mean O-AUROC/P-AUROC of 90.8%/96.0% and 95.9%/95.3% would move the state of the art by 8.5/7.8 points on Real3D-AD and by 0.6/1.0 points on MVTec3D-AD.

Reading between the lines

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

  • Because the transport is learned on one class of normal features, an immediate extension is to test the terminal Gaussian energy as a zero-shot out-of-distribution score on anomaly types never seen during training; the paper does not report this experiment.
  • The view-count ablation shows performance peaking at 27 views and declining at 64, so an adaptive view-selection or attention-based view weighting could push scores higher while saving computation; this is not in the paper.
  • The fixed nonlinearity in Eq. (13) is only one possible reading of the terminal state; replacing it with a calibrated likelihood of the endpoint under $\mathcal{N}(0,I)$ or a small learned normality head is a direct testable variant that might improve ranking on categories where detection lags.
  • The bidirectional projection recipe is not tied to point clouds for defect detection: any irregular 3D data that can be rendered into regular views—LiDAR sweeps, depth images, or partial scans—could inherit the same density-proxy transport.
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

3 major / 5 minor

Summary. The paper proposes MVFM-3DAD, an unsupervised 3D anomaly detection and localization method. It renders an input point cloud into multiple 2D views through a Bidirectional Geometric Projector (BGP), extracts features with a frozen DINOv2 ViT, and learns a flow-matching velocity field that transports normal multi-view features to a standard Gaussian reference through the Flow-guided Density Proxy Estimator (FDPE). At inference, the terminal Gaussian energy of an ODE-integrated test feature is converted via Eq. (13) into per-pixel anomaly evidence, which is back-projected to 3D points and pooled to an object-level score. The method is evaluated on Real3D-AD and MVTec3D-AD, reporting mean O-AUROC/P-AUROC of 90.8%/96.0% and 95.9%/95.3%, respectively.

Significance. If the reported results are reliable, MVFM-3DAD would be a meaningful advance: it offers a density-oriented alternative to memory-bank and reconstruction-based 3DAD, avoids explicit Jacobian computation, and shows large margins on Real3D-AD, especially for point-level localization. The paper includes a useful paradigm-level ablation and a hyperparameter analysis, and the code is promised. The principal weaknesses are the heuristic, unverified OOD transport assumption underlying Eq. (13) and the small, non-repeated MVTec3D-AD margins, both of which currently leave the central claim partially unsupported.

major comments (3)
  1. [Section 3.3, Eqs. (9)-(13)] The anomaly score is the terminal Gaussian energy of a feature transported by a velocity field trained only on normal features. The paper provides no derivation or bound showing that anomalous features land in the high-energy region of the reference Gaussian; the only supporting statement is that the terminal state is "expected to be compatible" after Eq. (13). For normal features the construction holds by training, but for anomalous inputs the terminal distribution is unconstrained, so nothing in the paper prevents an anomalous feature from being transported into the typical set of the Gaussian. Because this monotonicity property is the load-bearing premise for the reported AUROC values, the authors should either prove a suitable OOD transport bound under stated assumptions or directly verify the ordering empirically (for example, with per-category histograms of normal versus anomalous terminal energies, and with the score-AUROC as a function of ODE steps). As written, the central mechanism is an unverified heuristic.
  2. [Section 4.1 and Table 2] All experiments use a fixed random seed of 1 with a single run. The MVTec3D-AD advantage over HGCF is only 0.6% in O-AUROC (95.9 versus 95.3), and the P-AUROC margin over the best baseline is also small (95.3 versus 93.4 for LSFA, not the 1.0% stated in Section 4.3). Without repeated runs, standard deviations, or significance tests, the claim that MVFM-3DAD outperforms the strongest competitors on MVTec3D-AD is not statistically supported. Please report results over at least several seeds with means and standard deviations, and preferably a paired significance test.
  3. [Table 3] The ablation compares BGP+FDPE with memory-bank and reconstruction variants, but it does not isolate the contribution of the flow-guided terminal-energy score. A control experiment that replaces the transported terminal energy in Eq. (13) with an equally simple feature-space score, such as the squared norm of the raw normalized feature or the distance to a class-mean feature, would show whether the gain comes from the ODE transport or from the multi-view representation and back-projection alone. As written, the attribution of the reported improvement to "density proxy estimation" is not established.
minor comments (5)
  1. [Section 4.1] The list of baselines contains 'MVR-PCLIP' twice; the duplicate should be removed.
  2. [Sections 4.2 and Table 1] The text states that MVFM-3DAD ranks first in 9 of 12 O-AUROC categories, but the table shows first place in 10 of 12 categories, with Airplane and Duck being the exceptions; please correct or clarify the counting.
  3. [Fig. 4] The axis labels 'O-ROC' and 'P-ROC' should be 'O-AUROC' and 'P-AUROC' for consistency with the rest of the paper.
  4. [Table 2] Several entries contain formatting errors, including '90.4./99.2' and '86.8./97.1' for PointAD and a stray comma in '58.7/93.6,' for AnomalyCLIP; these should be fixed.
  5. [Sections 2.2 and 3.3] The FDPE is described as following the density-proxy perspective of WT-Flow [11], but the specific new aspects relative to [11], beyond the BGP projection and back-projection, should be stated explicitly; as written, the novelty appears to reside mainly in the BGP, and the paper should clarify the exact difference in the density-proxy mechanism.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the density-proxy criterion is adopted as a stated assumption, and the reported improvements are measured on independent benchmarks.

full rationale

MVFM-3DAD's derivation chain is self-contained with respect to its empirical claims. The FDPE training loss (Eq. 9) fits a velocity field to normal training features, and the anomaly evidence (Eq. 13) is a fixed deterministic function of the ODE terminal state; no parameter is fitted to test labels or to the reported AUROC values. The BGP forward/backward mapping (Eqs. 3 and 5) is a deterministic geometric correspondence transfer, and the object score (Eq. 14) is simply the maximum point response. The paper does not claim to derive the OOD transport property that anomalous features land in high terminal Gaussian energy; it explicitly says the terminal state is 'expected to be compatible' and adopts the density-proxy perspective from WT-Flow [11]. That is a self-citation for design provenance, and the underlying assumption is a correctness or robustness risk rather than a circularity: the success of the method is checked against Real3D-AD and MVTec3D-AD, which are independent of the fitted values and of the cited prior work. No equation in the paper reduces by construction to a fitted value, a quantity defined by the target result, or a self-citation chain that forces the reported outcome.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method adds no new physical entities and its core scoring formula contains no fitted constants beyond the hyperparameters listed. It rests on strong domain assumptions about multi-view rendering, OOD transport behavior, and projection fidelity, plus two hyperparameters tuned on the Real3D-AD benchmark.

free parameters (2)
  • Number of rendered views V = 27
    Selected by hyperparameter analysis on Real3D-AD (Fig. 4); performance peaks at 27 views and declines at 64, so this value is tuned to the evaluation benchmark.
  • Number of ODE steps = 20
    Selected from hyperparameter analysis on Real3D-AD (Fig. 4); performance stabilizes after 20 steps. Tuning on the test benchmark can inflate apparent performance, and no separate validation set is described.
assumptions (4)
  • domain assumption Rendered multi-view images preserve the information needed to detect and localize 3D anomalies in the frozen DINOv2 feature space.
    The whole pipeline projects 3D to 2D and relies on 2D features; if occlusion or rendering loses defect signal, point-level scoring fails. Entered in Section 3.2.
  • domain assumption A flow-matching velocity field trained only on normal features will transport abnormal features to high terminal Gaussian energy, so Eq. (13) is a monotone proxy for anomaly-ness.
    No theoretical guarantee or calibration is given; the method's discriminative power depends on this OOD transport behavior. Entered in Section 3.3, Eqs. (10)-(13).
  • domain assumption Bilinear upsampling and sampling at stored projection coordinates accurately transfer feature-level evidence back to point-level scores across all views.
    Eq. (5) assumes coordinate correspondences remain valid after upsampling and that averaging over views preserves localization; occluded points rely on visible views. Entered in Section 3.2.
  • standard math The conditional flow-matching objective in Eq. (9) with the linear path in Eq. (7) is a valid way to learn a probability path from the data distribution to a Gaussian, as established by WT-Flow and general flow-matching theory.
    The paper relies on this to treat terminal Gaussian energy as a density proxy; it is a known result rather than a new derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MVFM-3DAD: Multi-view Flow Matching for 3D Anomaly Detection via Density Proxy Estimation." pith.science (2026). https://pith.science/paper/TNRQLVWR

@misc{pith2026260812148,
  author       = {Pith},
  title        = {Pith review of: MVFM-3DAD: Multi-view Flow Matching for 3D Anomaly Detection via Density Proxy Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TNRQLVWR}},
  note         = {Machine review of arXiv:2608.12148}
}
read the original abstract

In 3D anomaly detection (3DAD), most existing methods rely on Memory bank retrieval or reconstruction. However, memory-based methods are constrained by the coverage of stored normal features, while reconstruction-based methods may learn identity shortcuts that also reconstruct anomalous inputs well. These limitations motivate a density-oriented approach that evaluates whether a test sample follows the learned normal distribution. To this end, we propose MVFM-3DAD, a flow-based framework that reframes 3DAD as density proxy estimation over the normal data distribution. MVFM-3DAD introduces a Bidirectional Geometric Projector (BGP), whose forward process converts irregular point clouds into structured multi-view representations. The Flow-guided Density Proxy Estimator (FDPE) estimates a reference density for each view feature, after which the backward process of BGP maps these multi-view density estimates to their corresponding 3D points. Building on it, anomalous features can be identified by their terminal normality. Unlike conventional flow-based likelihood estimation, our formulation requires neither input reconstruction nor explicit Jacobian evaluation, yielding a simple and efficient anomaly-scoring mechanism. Extensive experiments show that MVFM-3DAD outperforms the strongest competing methods on Real3D-AD and MVTec3D-AD. Code is available at https://github.com/lil-wayne-0319/MV3D-AD

Figures

Figures reproduced from arXiv: 2608.12148 by the authors.

Figure 1
Figure 1. Overview of MVFM-3DAD. BGP projects 3D point clouds into multi-view images and maps anomaly evidence back to 3D points, while FDPE estimates anomaly scores through flow-guided density proxy modeling. 3.2 Bidirectional Geometric Projector Direct density estimation within the raw 3D input space presents a formidable challenge, as the observations inherently resides on complex low-dimensional manifolds and contains tas… view at source ↗
Figure 2
Figure 2. Overview of the proposed FDPE module. Patch tokens are derived from a frozen pretrained feature extractor. Density-proxy inference. Given a test view feature F v i , we initialize the ODE state by the normalized feature: Z v i,0 = LN (F v i ). (10) The learned transport is then solved from t = 0 to t = 1: dZ v i,t dt = gψ [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results on the Real3D-AD and MVTec3D-AD benchmark [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Hyperparameter analysis on the Real3D-AD benchmark. 4.6 Analysis of hyperparameters We quantitatively analyze the impact of two key parameters in our framework, the number of views rendered by the BGP and the inference ODE steps utilized in the FDPE. The results in [P…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [1]

    arXiv preprint arXiv:2112.09045 (2021)

    Bergmann, P., Jin, X., Sattlegger, D., Steger, C.: The mvtec 3d-ad dataset for un- supervised 3d anomaly detection and localization. arXiv preprint arXiv:2112.09045 (2021)

    arXiv 2021

  2. [2]

    Pattern Recog

    Cao, Y., Xu, X., Shen, W.: Complementary pseudo multimodal feature for point cloud anomaly detection. Pattern Recog. (2024)

    work page 2024

  3. [3]

    In: ACM Int

    Chen, R., Xie, G., Liu, J., Wang, J., Luo, Z., Wang, J., Zheng, F.: Easynet: An easy network for 3d industrial anomaly detection. In: ACM Int. Conf. Multimedia (2023)

    work page 2023

  4. [4]

    Flow Mismatching: Unsupervised Anomaly Detection via Velocity Discrepancies in Flow Matching Models

    Chen, S., Moradi, M., Paynabar, K., Yan, H.: Flow mismatching: Unsupervised anomaly detection via velocity discrepancies in flow matching models. arXiv preprint arXiv:2605.23070 (2026)

    work page Pith review arXiv 2026

  5. [5]

    IEEE Transactions on Sys- tems, Man, and Cybernetics: Systems (2025)

    Cheng, Y., Cao, Y., Xie, G., Lu, Z., Shen, W.: Toward zero-shot point cloud anomaly detection: A multiview projection framework. IEEE Transactions on Sys- tems, Man, and Cybernetics: Systems (2025)

    work page 2025

  6. [6]

    Chu, Y.M., Liu, C., Hsieh, T.I., Chen, H.T., Liu, T.L.: Shape-guided dual-memory learning for 3d anomaly detection. In: Int. Conf. Mach. Learn. PMLR (2023) 12 Authors Suppressed Due to Excessive Length

    work page 2023

  7. [7]

    In: Proc

    Gu, Z., Zhang, J., Liu, L., Chen, X., Peng, J., Gan, Z., Jiang, G., Shu, A., Wang, Y., Ma, L.: Rethinking reverse distillation for multi-modal anomaly detection. In: Proc. AAAI Conf. Artif. Intell. (2024)

    work page 2024

  8. [8]

    In: Proc

    Gudovskiy,D.,Ishizaka,S.,Kozuka,K.:Cflow-ad:Real-timeunsupervisedanomaly detection with localization via conditional normalizing flows. In: Proc. IEEE/CVF winter Conf. on Appl. of Comput. Vis. (2022)

    work page 2022

Show all 32 references
  1. [9]

    In: Proc

    Horwitz, E., Hoshen, Y.: Back to the feature: classical 3d features are (almost) all you need for 3d anomaly detection. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2023)

    2023

  2. [10]

    In: Proc

    Kim, S., Lee, W., Cho, M.: A semantically disentangled unified model for multi- category 3d anomaly detection. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2026)

    2026

  3. [11]

    arXiv preprint arXiv:2508.05461 (2025)

    Li, L., Liu, L., Liu, J., Zhang, J., Hao, R., Du, X.: How and why: Taming flow matching for unsupervised anomaly detection and localization. arXiv preprint arXiv:2508.05461 (2025)

    2025 arXiv

  4. [12]

    Li, M., He, J., Li, J., Han, D., Wan, J., Li, G.: Hgcf: Hierarchical geometry-color fusionformultimodalindustrialanomalydetection.In:ACMInt.Conf.Multimedia (2025)

    2025

  5. [13]

    In: Proc

    Li, W., Xu, X., Gu, Y., Zheng, B., Gao, S., Wu, Y.: Towards scalable 3d anomaly detection and localization: A benchmark via 3d anomaly synthesis and a self- supervised learning network. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2024)

    2024

  6. [14]

    Li, Z., Huang, Q., Zhu, Y., Yang, L., Mohammadi Amiri, M., van Stein, N., van Leeuwen, M.: Scalable, explainable and provably robust anomaly detection with one-step flow matching. Adv. Neural Inform. Process. Syst. (2026)

    2026

  7. [15]

    Neural Networks p

    Liang, H., Hu, C., Tang, Y., Shen, L., Wang, J., Gao, C.: Mff-m3ad: A unified re- construction method with multi-scale feature fusion for multi-category 3d anomaly detection. Neural Networks p. 109131 (2026)

    2026

  8. [16]

    In: Proc

    Liang, H., Xie, G., Hou, C., Wang, B., Gao, C., Wang, J.: Look inside for more: Internal spatial modality perception for 3d anomaly detection. In: Proc. AAAI Conf. Artif. Intell. (2025)

    2025

  9. [17]

    In: ACM Int

    Liang, H., Zhang, J., Dai, T., Shen, L., Wang, J., Gao, C.: Taming anomalies with down-up sampling networks: Group center preserving reconstruction for 3d anomaly detection. In: ACM Int. Conf. Multimedia (2025)

    2025

  10. [18]

    Pattern Recog- nition p

    Liang, H., Zhou, J., Gao, C., Guo, B., Wang, J., Shen, L.: A lightweight 3d anomaly detection method with rotationally invariant features. Pattern Recog- nition p. 112924 (2025)

    2025

  11. [19]

    Liu, J., Xie, G., Chen, R., Li, X., Wang, J., Liu, Y., Wang, C., Zheng, F.: Real3d- ad: A dataset of point cloud anomaly detection. Adv. Neural Inform. Process. Syst. (2023)

    2023

  12. [20]

    In: Proc

    Long, K., Xie, G., Ma, L., Liu, J., Lu, Z.: Revisiting multimodal fusion for 3d anomaly detection from an architectural perspective. In: Proc. AAAI Conf. Artif. Intell. (2025)

    2025

  13. [21]

    In: Proc

    Roth, K., Pemula, L., Zepeda, J., Scholkopf, B., Brox, T., Gehler, P.: Towards total recall in industrial anomaly detection. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2022)

    2022

  14. [22]

    In: Proc

    Rudolph, M., Wehrbein, T., Rosenhahn, B., Wandt, B.: Asymmetric student- teacher networks for industrial anomaly detection. In: Proc. IEEE/CVF winter Conf. on Appl. of Comput. Vis. (2023) Title Suppressed Due to Excessive Length 13

    2023

  15. [23]

    In: Proc

    Sakai, S., He, X., Gu, C., Sigal, L., Hasegawa, T.: Invad: Inversion-based reconstruction-free anomaly detection with diffusion models. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2026)

    2026

  16. [24]

    IEEE Trans

    Wang, C., Zhu, H., Peng, J., Wang, Y., Yi, R., Wu, Y., Ma, L., Zhang, J.: M3dm- nr: Rgb-3d noisy-resistant industrial anomaly detection via multimodal denoising. IEEE Trans. Pattern Anal. Mach. Intell. (2025)

    2025

  17. [25]

    In: Proc

    Wu, L., Zou, Q., Yue, Y., Wang, Z.: Geometry-aligned and anomaly-aware re- construction for 3d anomaly detection. In: Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. (2026)

    2026

  18. [26]

    arXiv preprint arXiv:2111.07677 (2021)

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

    2021 arXiv

  19. [27]

    In: Proc

    Zavrtanik, V., Kristan, M., Skočaj, D.: Cheating depth: Enhancing 3d surface anomaly detection via depth simulation. In: Proc. IEEE/CVF winter Conf. on Appl. of Comput. Vis. (2024)

    2024

  20. [28]

    In: Proc

    Zha, Y., Yuerong, X., Fan, C., Wang, Y., Dai, T., Chen, K., Xia, S.T.: Casl: Curvature-augmented self-supervised learning for 3d anomaly detection. In: Proc. AAAI Conf. Artif. Intell. (2026)

    2026

  21. [29]

    Zhou, Q., Pang, G., Tian, Y., He, S., Chen, J.: Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection. In: Int. Conf. Learn. Represent. (2024)

    2024

  22. [30]

    Zhou, Q., Yan, J., He, S., Meng, W., Chen, J.: Pointad: Comprehending 3d anoma- lies from points and pixels for zero-shot 3d anomaly detection. Adv. Neural Inform. Process. Syst. (2024)

    2024

  23. [31]

    IEEE Trans

    Zhou, Y., Xu, X., Song, J., Shen, F., Shen, H.T.: Msflow: Multiscale flow-based framework for unsupervised anomaly detection. IEEE Trans. Neural Netw. Learn. Syst. (2024)

    2024

  24. [32]

    Zhou, Z., Wang, L., Fang, N., Wang, Z., Qiu, L., Zhang, S.: R3d-ad: Reconstruction via diffusion for 3d anomaly detection. In: Eur. Conf. Comput. Vis. Springer (2024)

    2024

Pith tools

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