Pith. sign in

REVIEW 3 major objections 58 references

Pretrained point-language models can detect and localize 3D industrial defects without any target-category training data.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review

2026-07-14 20:17 UTC pith:GJIDBRGC

load-bearing objection Solid first PLM-native ZS 3D AD system with real point-level gains; abstract overclaims overall superiority and the ZS framing is cross-category, not label-free. the 3 major comments →

arxiv 2603.21511 v3 pith:GJIDBRGC submitted 2026-03-23 cs.CV

Back to Point: Exploring Point-Language Models for Zero-Shot 3D Anomaly Detection

Kaiqiang Li , Gang Li , Mingle Zhou , Min Li , Delong Han , Jin Wan This is my paper
classification cs.CV
keywords zero-shot 3D anomaly detectionpoint-language modelsmulti-granularity feature embeddinggeometric descriptorsindustrial inspectionpoint cloud anomaly localizationjoint representation learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Zero-shot 3D anomaly detection matters for industrial inspection when target defect samples cannot be collected. Existing methods convert point clouds into multi-view images and run vision-language models, which discard geometry and depend on view choice. This paper argues that pretrained point-language models can work directly on the point cloud instead. It introduces BTP, which aligns multi-granularity patch features and geometric descriptors with normal versus defective text embeddings, and trains with a joint loss on auxiliary point-cloud data. On Real3D-AD and Anomaly-ShapeNet the method reports stronger point-level localization than prior zero-shot approaches and several supervised baselines, while remaining competitive at object level.

Core claim

Zero-shot 3D anomaly detection and localization can be performed directly in point-cloud space by aligning multi-granularity patch features, a global CLS token, and learnable geometric descriptors from a pretrained point-language encoder with hybrid normal and defective text prompts, without target-category training data or 2D rendering.

What carries the argument

Multi-Granularity Feature Embedding Module (MGFEM): fuses intermediate encoder patch features, the CLS token, and PointNet-based geometric descriptors into a shared text-aligned space so cosine similarity to normal versus defective prompts yields point-level anomaly scores; jointly optimized with global, local (focal plus dice), and geometric (contrastive to FPFH) losses.

Load-bearing premise

That training on labeled anomalies from other categories, plus similarity between fused point-language embeddings and normal-or-defective text prompts, transfers to completely unseen industrial categories with no target samples.

What would settle it

Train BTP on Real3D-AD categories that share little structure with a held-out category (for example omit gem-like objects and test on diamond) and check whether point-level AUROC falls below competitive multi-view zero-shot baselines.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Inspection systems can localize defects on unseen object categories without collecting target anomalous samples.
  • Multi-view rendering and back-projection are unnecessary for language-guided 3D anomaly detection, cutting view bias and projection cost.
  • Intermediate patch features from pretrained point-language models, not only global embeddings, carry transferable anomaly-sensitive structure.
  • Joint geometric, local, and global supervision on auxiliary categories strengthens zero-shot point-level scores.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Object-level scores lag point-level ones, so better aggregation of local anomalies into a global decision may close the gap the paper itself flags.
  • The same multi-granularity text alignment may extend to other fine-grained zero-shot point-cloud tasks such as part-level damage grading.
  • If the learnable geometric branch fully absorbs handcrafted FPFH priors, inference could drop those descriptors while keeping the training alignment.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 0 minor

Summary. The paper proposes BTP, a zero-shot 3D anomaly detection framework that operates directly on point clouds with a pretrained Point-Language Model (ULIP/ULIP2) rather than multi-view VLM rendering. It extracts multi-layer patch features, a CLS token, and learnable geometric descriptors (GFCM, PointNet-style, contrastively aligned to FPFH), fuses them in MGFEM into text-aligned patch embeddings, and uses hybrid learnable normal/defective prompts. Joint training combines focal+dice point-level losses, object-level BCE, and geometric contrastive loss on auxiliary categories under a one-class cross-category protocol. On Real3D-AD and Anomaly-ShapeNet, BTP reports strong point-level localization (mean P-AUROC 84.5% and 87.3%) and competitive but weaker object-level scores, with module/loss and point-count ablations.

Significance. If the localization gains hold under a clearly stated protocol, this is a useful contribution: it is among the first systematic attempts to use pretrained PLMs for ZS 3D anomaly detection without 3D-to-2D projection, preserves geometry, and shows large point-level improvements over projection-based ZS baselines and several supervised/unsupervised methods. The multi-granularity fusion, geometry branch, and joint losses are concrete and ablated; code is promised. The work is practically relevant for industrial inspection where target-category data are scarce, even if object-level scoring remains weaker than PointAD.

major comments (3)
  1. Abstract, Introduction, and Sec. 2.1 frame BTP as zero-shot detection “without anomalous samples” / “without requiring any target-category training data,” while Sec. 3.5–3.6 and Eqs. (5)–(8) train with point-level anomaly masks (focal+dice), object labels (BCE), and FPFH contrastive loss on auxiliary categories (Tab. 2 one-class cross-category protocol). This is cross-category transfer with anomaly labels, not a no-anomaly-label or pure prompt-only ZS setting like the VLM baselines it is compared to. The abstract and claims should be restated to match the actual protocol, and a true no-anomaly-label (or frozen-encoder, prompt-only) baseline should be reported so the localization gains are not overstated relative to PointAD/AnomalyCLIP.
  2. Table 1 and Tab. 3: object-level O-AUROC is 61.4% vs PointAD 74.8% (and below several supervised methods), while point-level is clearly best (84.5% P-AUROC, 81.9% P-PRO). The abstract’s blanket “superior performance in ZS 3D anomaly detection” is not supported for object-level detection. Either qualify the claim to localization, or strengthen object-level scoring/aggregation (the conclusion itself flags this gap). Without that, the central “superior ZS” claim is only half-supported.
  3. Sec. 3.3–3.4 and Eq. (4): MGFEM is load-bearing for the localization claim, but the fusion is under-specified (projection layers, softmax layer weights α_l, how multi-layer patches are pooled to points, and how patch scores become point scores). Reproducibility and the claim that multi-granularity alignment drives the +8.6 P-AUROC gain over CPMF require a clearer formal definition and, ideally, a short sensitivity check on which intermediate layers (stated as layers 4–11 in Sec. 4.2) matter.

Circularity Check

0 steps flagged

No circularity: empirical PLM adaptation evaluated on held-out categories and public benchmarks; metrics are not forced by construction.

full rationale

BTP is a standard empirical computer-vision system paper. Its core claims (multi-granularity patch-text alignment via MGFEM, GFCM geometric descriptors, hybrid prompts, and the joint loss of Eqs. 5–8) are architectural and training choices, not first-principles derivations. Anomaly scores are cosine similarities between fused ULIP embeddings and text prompts; reported O-/P-AUROC, AP, PRO, etc. are computed on held-out categories under the one-class cross-category protocol (Tab. 1–3) and public datasets (Real3D-AD, Anomaly-ShapeNet). No equation reduces a claimed prediction to a fitted constant by identity; loss weights and prompt tokens are optimized but the evaluation metrics remain external. Self-citations (e.g., related MCL-AD work) appear only in Related Work and are not load-bearing for uniqueness or for the numerical results. Terminology debates about the strictness of “zero-shot” (auxiliary anomaly labels are used) concern experimental protocol, not circular derivation. The paper is therefore self-contained against its benchmarks; score 0 with empty steps is the correct outcome.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 3 invented entities

Load-bearing content is mostly standard deep-learning practice plus the assumption that ULIP’s shared point-text space and auxiliary-category anomaly labels suffice for cross-category ZS localization. Free parameters are ordinary hyperparameters; invented pieces are architectural modules validated only inside this paper’s tables.

free parameters (5)
  • loss weights λ_global / λ_geo (and local combination) = 0.5 and 0.5 (stated)
    Eq. (8) balances local, global, and geometric losses; paper sets coefficients to 0.5/0.5 (and focal+dice for local). Central metrics depend on this hand-chosen mix.
  • learnable text prompt length = 4
    Hybrid prompt learner length set to 4; affects normal/abnormal text embeddings used for scoring.
  • intermediate encoder layers for multi-layer semantics = layers as in Sec. 4.2 / [38]
    Patch features taken from chosen ULIP/Point-BERT layers (following PLANE-style extraction); which layers are used is a design choice that drives localization.
  • input point count (FPS) = 2048
    Default 2048 points chosen after ablation; changes P-AUROC and speed (Tab. 5).
  • object-level fusion coefficient α
    Eq. (6) mixes point- and patch-level predictions with balancing coefficient α for BCE object score.
axioms (4)
  • domain assumption Cosine similarity between point/patch embeddings and normal vs abnormal text embeddings is a valid anomaly score in the ULIP-aligned space.
    Core of Sec. 3.1–3.2 and scoring pipeline; inherited from CLIP-style ZS AD without independent proof for industrial 3D defects.
  • domain assumption Training on labeled anomalies from non-target categories yields zero-shot generalization to unseen industrial categories.
    One-class cross-category protocol in Sec. 4.3 and Tab. 2; standard ZS AD framing but load-bearing for the claim of no target-category data.
  • ad hoc to paper Learnable PointNet patch descriptors can absorb FPFH geometric priors via contrastive alignment and improve structural anomaly sensitivity.
    GFCM + L_geo (Eq. 2, 7); justified by ablation, not by external theorem.
  • standard math Standard optimization and metric definitions (AUROC, AP, F1, AU-PRO) correctly rank detectors for industrial use.
    Sec. 4.1 evaluation protocol.
invented entities (3)
  • BTP framework (MGFEM + GFCM + hybrid prompts + joint losses) no independent evidence
    purpose: End-to-end zero-shot point-level and object-level 3D anomaly scoring without multi-view rendering.
    Composite system introduced here; evidence is internal benchmarks only.
  • Multi-Granularity Feature Embedding Module (MGFEM) no independent evidence
    purpose: Fuse multi-layer patch semantics, CLS token, and geometric descriptors into text-aligned patch embeddings.
    Sec. 3.4; performance gains shown only in this paper’s ablations.
  • Geometric Feature Creation Module (GFCM) no independent evidence
    purpose: Learnable PointNet replacement for handcrafted FPFH patch geometry.
    Sec. 3.3; independent_evidence false beyond FPFH contrastive target already known.

pith-pipeline@v1.1.0-grok45 · 21122 in / 3364 out tokens · 36128 ms · 2026-07-14T20:17:09.169356+00:00 · methodology

0 comments
read the original abstract

Zero-shot (ZS) 3D anomaly detection is crucial for reliable industrial inspection, as it enables detecting and localizing defects without requiring any target-category training data. Existing approaches render 3D point clouds into 2D images and leverage pre-trained Vision-Language Models (VLMs) for anomaly detection. However, such strategies inevitably discard geometric details and exhibit limited sensitivity to local anomalies. In this paper, we revisit intrinsic 3D representations and explore the potential of pre-trained Point-Language Models (PLMs) for ZS 3D anomaly detection. We propose BTP (Back To Point), a novel framework that effectively aligns 3D point cloud and textual embeddings. Specifically, BTP aligns multi-granularity patch features with textual representations for localized anomaly detection, while incorporating geometric descriptors to enhance sensitivity to structural anomalies. Furthermore, we introduce a joint representation learning strategy that leverages auxiliary point cloud data to improve robustness and enrich anomaly semantics. Extensive experiments on Real3D-AD and Anomaly-ShapeNet demonstrate that BTP achieves superior performance in ZS 3D anomaly detection. Code will be available at \href{https://github.com/wistful-8029/BTP-3DAD}{https://github.com/wistful-8029/BTP-3DAD}.

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 2 linked inside Pith

  1. [1]

    Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection

    Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. Mvtec ad–a comprehensive real-world dataset for unsupervised anomaly detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9592–9600, 2019. 1

    2019

  2. [2]

    Comple- mentary pseudo multimodal feature for point cloud anomaly detection.Pattern Recognition, 156:110761, 2024

    Yunkang Cao, Xiaohao Xu, and Weiming Shen. Comple- mentary pseudo multimodal feature for point cloud anomaly detection.Pattern Recognition, 156:110761, 2024. 1, 2, 6

    2024

  3. [3]

    Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly de- tection

    Yunkang Cao, Jiangning Zhang, Luca Frittoli, Yuqi Cheng, Weiming Shen, and Giacomo Boracchi. Adaclip: Adapting clip with hybrid learnable prompts for zero-shot anomaly de- tection. InEuropean Conference on Computer Vision, pages 55–72. Springer, 2024. 2

    2024

  4. [4]

    Personalizing vision- language models with hybrid prompts for zero-shot anomaly detection.IEEE Transactions on Cybernetics, 2025

    Yunkang Cao, Xiaohao Xu, Yuqi Cheng, Chen Sun, Zongwei Du, Liang Gao, and Weiming Shen. Personalizing vision- language models with hybrid prompts for zero-shot anomaly detection.IEEE Transactions on Cybernetics, 2025. 2

    2025

  5. [5]

    Shapenet: An information-rich 3d model repository.arXiv preprint arXiv:1512.03012, 2015

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, et al. Shapenet: An information-rich 3d model repository.arXiv preprint arXiv:1512.03012, 2015. 5

    Pith/arXiv arXiv 2015

  6. [6]

    Dis- tilled large language model-driven dynamic sparse expert ac- tivation mechanism.Applied Soft Computing, page 114037,

    Qinghui Chen, Zekai Zhang, Zaigui Zhang, Kai Zhang, Da- gang Li, Wenmin Wang, Jinglin Zhang, and Cong Liu. Dis- tilled large language model-driven dynamic sparse expert ac- tivation mechanism.Applied Soft Computing, page 114037,

  7. [7]

    Easynet: An easy net- work for 3d industrial anomaly detection

    Ruitao Chen, Guoyang Xie, Jiaqi Liu, Jinbao Wang, Ziqi Luo, Jinfan Wang, and Feng Zheng. Easynet: An easy net- work for 3d industrial anomaly detection. InProceedings of the 31st ACM International Conference on Multimedia, pages 7038–7046, 2023. 1

    2023

  8. [8]

    Toward zero-shot point cloud anomaly de- tection: A multiview projection framework.IEEE Transac- tions on Systems, Man, and Cybernetics: Systems, 2025

    Yuqi Cheng, Yunkang Cao, Guoyang Xie, Zhichao Lu, and Weiming Shen. Toward zero-shot point cloud anomaly de- tection: A multiview projection framework.IEEE Transac- tions on Systems, Man, and Cybernetics: Systems, 2025. 2

    2025

  9. [9]

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

    Yu-Min Chu, Chieh Liu, Ting-I Hsieh, Hwann-Tzong Chen, and Tyng-Luh Liu. Shape-guided dual-memory learning for 3d anomaly detection. InProceedings of the 40th Interna- tional Conference on Machine Learning, pages 6185–6194,

  10. [10]

    3d scanning method for robotized inspection of industrial sealed parts

    Kleber Roberto da Silva Santos, Wesley Rodrigues de Oliveira, Em´ılia Villani, and Augusto Dttmann. 3d scanning method for robotized inspection of industrial sealed parts. Computers in Industry, 147:103850, 2023. 1

    2023

  11. [11]

    Bert: Pre-training of deep bidirectional trans- formers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional trans- formers for language understanding. InProceedings of the 2019 conference of the North American chapter of the asso- ciation for computational linguistics: human language tech- nologies, volume 1 (long and short papers), pages 4171– 4186, 2019. 3

    2019

  12. [12]

    3d octave and 2d vanilla mixed convolutional neural network for hyperspectral image classification with limited samples.Remote Sensing, 13(21):4407, 2021

    Yuchao Feng, Jianwei Zheng, Mengjie Qin, Cong Bai, and Jinglin Zhang. 3d octave and 2d vanilla mixed convolutional neural network for hyperspectral image classification with limited samples.Remote Sensing, 13(21):4407, 2021. 2

    2021

  13. [13]

    A comprehensive performance evaluation of 3d local feature descriptors.Inter- national Journal of Computer Vision, 116(1):66–89, 2016

    Yulan Guo, Mohammed Bennamoun, Ferdous Sohel, Min Lu, Jianwei Wan, and Ngai Ming Kwok. A comprehensive performance evaluation of 3d local feature descriptors.Inter- national Journal of Computer Vision, 116(1):66–89, 2016. 3

    2016

  14. [14]

    Deep learning for 3d point clouds: A survey.IEEE transactions on pattern analysis and machine intelligence, 43(12):4338–4364, 2020

    Yulan Guo, Hanyun Wang, Qingyong Hu, Hao Liu, Li Liu, and Mohammed Bennamoun. Deep learning for 3d point clouds: A survey.IEEE transactions on pattern analysis and machine intelligence, 43(12):4338–4364, 2020. 3

    2020

  15. [15]

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

    Eliahu Horwitz and Yedid Hoshen. Back to the feature: clas- sical 3d features are (almost) all you need for 3d anomaly detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2968– 2977, 2023. 1, 6, 7

    2023

  16. [16]

    Research on product surface quality in- spection technology based on 3d point cloud.Advances in Mechanical Engineering, 15(3):16878132231159523, 2023

    Lintao Huo, Ying Liu, Yutu Yang, Zilong Zhuang, and Mengmeng Sun. Research on product surface quality in- spection technology based on 3d point cloud.Advances in Mechanical Engineering, 15(3):16878132231159523, 2023. 1

    2023

  17. [17]

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

    Jongheon Jeong, Yang Zou, Taewan Kim, Dongqing Zhang, Avinash Ravichandran, and Onkar Dabeer. Winclip: Zero- /few-shot anomaly classification and segmentation. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19606–19616, 2023. 2

    2023

  18. [18]

    Using spin images for efficient object recognition in cluttered 3d scenes.IEEE Transactions on pattern analysis and machine intelligence, 21(5):433–449, 2002

    Andrew E Johnson and Martial Hebert. Using spin images for efficient object recognition in cluttered 3d scenes.IEEE Transactions on pattern analysis and machine intelligence, 21(5):433–449, 2002. 3

    2002

  19. [19]

    3dffl: privacy- preserving federated few-shot learning for 3d point clouds in autonomous vehicles.Scientific Reports, 14(1):19589, 2024

    Abdullah Aman Khan, Khwaja Mutahir Ahmad, Sidra Shafiq, Waqas Amin, and Rajesh Kumar. 3dffl: privacy- preserving federated few-shot learning for 3d point clouds in autonomous vehicles.Scientific Reports, 14(1):19589, 2024. 1

    2024

  20. [20]

    Mcl-ad: Multimodal collaboration learn- ing for zero-shot 3d anomaly detection.arXiv preprint arXiv:2509.10282, 2025

    Gang Li, Tianjiao Chen, Mingle Zhou, Min Li, Delong Han, and Jin Wan. Mcl-ad: Multimodal collaboration learn- ing for zero-shot 3d anomaly detection.arXiv preprint arXiv:2509.10282, 2025. 2

    Pith/arXiv arXiv 2025

  21. [21]

    Towards scalable 3d anomaly detection and localization: A benchmark via 3d anomaly synthesis and a self-supervised learning network

    Wenqiao Li, Xiaohao Xu, Yao Gu, Bozhong Zheng, Shenghua Gao, and Yingna Wu. Towards scalable 3d anomaly detection and localization: A benchmark via 3d anomaly synthesis and a self-supervised learning network. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 22207–22216, 2024. 1, 2, 5, 7

    2024

  22. [22]

    Multi-sensor object anomaly detection: Unifying appearance, geometry, and internal properties

    Wenqiao Li, Bozhong Zheng, Xiaohao Xu, Jinye Gan, Fad- ing Lu, Xiang Li, Na Ni, Zheng Tian, Xiaonan Huang, Shenghua Gao, et al. Multi-sensor object anomaly detection: Unifying appearance, geometry, and internal properties. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 9984–9993, 2025. 1

    2025

  23. [23]

    Real3d- ad: A dataset of point cloud anomaly detection.Advances in Neural Information Processing Systems, 36:30402–30415,

    Jiaqi Liu, Guoyang Xie, Ruitao Chen, Xinpeng Li, Jinbao Wang, Yong Liu, Chengjie Wang, and Feng Zheng. Real3d- ad: A dataset of point cloud anomaly detection.Advances in Neural Information Processing Systems, 36:30402–30415,

  24. [24]

    Openshape: Scaling up 3d shape representation towards open-world understanding.Advances in neural information processing systems, 36:44860–44879, 2023

    Minghua Liu, Ruoxi Shi, Kaiming Kuang, Yinhao Zhu, Xu- anlin Li, Shizhong Han, Hong Cai, Fatih Porikli, and Hao Su. Openshape: Scaling up 3d shape representation towards open-world understanding.Advances in neural information processing systems, 36:44860–44879, 2023. 3

    2023

  25. [25]

    Data-driven many-objective crowd worker selection for mobile crowdsourcing in indus- trial iot.IEEE Transactions on Industrial Informatics, 19(1): 531–540, 2021

    Zhuoran Lu, Yingjie Wang, Xiangrong Tong, Chunxiao Mu, Yu Chen, and Yingshu Li. Data-driven many-objective crowd worker selection for mobile crowdsourcing in indus- trial iot.IEEE Transactions on Industrial Informatics, 19(1): 531–540, 2021. 1

    2021

  26. [26]

    One-for-all few- shot anomaly detection via instance-induced prompt learn- ing

    Wenxi Lv, Qinliang Su, and Wenchao Xu. One-for-all few- shot anomaly detection via instance-induced prompt learn- ing. InThe Thirteenth International Conference on Learning Representations. 1

  27. [27]

    Aa-clip: Enhancing zero-shot anomaly detection via anomaly-aware clip

    Wenxin Ma, Xu Zhang, Qingsong Yao, Fenghe Tang, Chenxu Wu, Yingtai Li, Rui Yan, Zihang Jiang, and S Kevin Zhou. Aa-clip: Enhancing zero-shot anomaly detection via anomaly-aware clip. InProceedings of the Computer Vi- sion and Pattern Recognition Conference, pages 4744–4754,

  28. [28]

    Beyond single-modal boundary: Cross-modal anomaly detection through visual prototype and harmoniza- tion

    Kai Mao, Ping Wei, Yiyang Lian, Yangyang Wang, and Nan- ning Zheng. Beyond single-modal boundary: Cross-modal anomaly detection through visual prototype and harmoniza- tion. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 9964–9973, 2025. 2

    2025

  29. [29]

    Delving into out-of-distribution detection with vision-language representations

    Yifei Ming, Ziyang Cai, Jiuxiang Gu, Yiyou Sun, Wei Li, and Yixuan Li. Delving into out-of-distribution detection with vision-language representations. InAdvances in Neural Information Processing Systems, pages 35087–35102. Cur- ran Associates, Inc., 2022. 2

    2022

  30. [30]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 652–660,

  31. [31]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    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. 3

    2017

  32. [32]

    Bayesian prompt flow learning for zero-shot anomaly detec- tion

    Zhen Qu, Xian Tao, Xinyi Gong, Shichen Qu, Qiyu Chen, Zhengtao Zhang, Xingang Wang, and Guiguang Ding. Bayesian prompt flow learning for zero-shot anomaly detec- tion. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 30398–30408, 2025. 2

    2025

  33. [33]

    Learning transferable visual models from natural language supervi- sion

    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 supervi- sion. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021. 2

    2021

  34. [34]

    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. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14318–14328, 2022. 6, 7

    2022

  35. [35]

    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. 3

    2009

  36. [36]

    Shot: Unique signatures of histograms for surface and tex- ture description.Computer vision and image understanding, 125:251–264, 2014

    Samuele Salti, Federico Tombari, and Luigi Di Stefano. Shot: Unique signatures of histograms for surface and tex- ture description.Computer vision and image understanding, 125:251–264, 2014. 3

    2014

  37. [37]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017. 3

    2017

  38. [38]

    Exploiting point- language models with dual-prompts for 3d anomaly detec- tion.Expert Systems with Applications, page 129758, 2025

    Jiaxiang Wang, Haote Xu, Xiaolu Chen, Haodi Xu, Yue Huang, Xinghao Ding, and Xiaotong Tu. Exploiting point- language models with dual-prompts for 3d anomaly detec- tion.Expert Systems with Applications, page 129758, 2025. 1, 2, 6

    2025

  39. [39]

    Toward ac- curate anomaly detection in industrial internet of things us- ing hierarchical federated learning.IEEE Internet of Things Journal, 9(10):7110–7119, 2021

    Xiaoding Wang, Sahil Garg, Hui Lin, Jia Hu, Georges Kad- doum, Md Jalil Piran, and M Shamim Hossain. Toward ac- curate anomaly detection in industrial internet of things us- ing hierarchical federated learning.IEEE Internet of Things Journal, 9(10):7110–7119, 2021. 1

    2021

  40. [40]

    Dynamic graph cnn for learning on point clouds.ACM Transactions on Graphics (tog), 38(5):1–12, 2019

    Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E Sarma, Michael M Bronstein, and Justin M Solomon. Dynamic graph cnn for learning on point clouds.ACM Transactions on Graphics (tog), 38(5):1–12, 2019. 3

    2019

  41. [41]

    Multimodal industrial anomaly detection via hybrid fusion

    Yue Wang, Jinlong Peng, Jiangning Zhang, Ran Yi, Yabiao Wang, and Chengjie Wang. Multimodal industrial anomaly detection via hybrid fusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8032–8041, 2023. 6, 7

    2023

  42. [42]

    Multi- space crowd sensing task allocation: A dynamic co- optimization framework with fairness-aware reinforcement learning.IEEE Transactions on Mobile Computing, 2025

    Yingjie Wang, Dihong Luo, Haojun Teng, Peiyong Duan, Yang Gao, Haijing Zhang, and Zhipeng Cai. Multi- space crowd sensing task allocation: A dynamic co- optimization framework with fairness-aware reinforcement learning.IEEE Transactions on Mobile Computing, 2025. 1

    2025

  43. [43]

    Towards zero-shot 3d anomaly localization

    Yizhou Wang, Kuan-Chuan Peng, and Yun Fu. Towards zero-shot 3d anomaly localization. In2025 IEEE/CVF Win- ter Conference on Applications of Computer Vision (WACV), pages 1447–1456. IEEE, 2025. 2

    2025

  44. [44]

    Towards zero-shot anomaly detection and reasoning with multimodal large language models

    Jiacong Xu, Shao-Yuan Lo, Bardia Safaei, Vishal M Patel, and Isht Dwivedi. Towards zero-shot anomaly detection and reasoning with multimodal large language models. InPro- ceedings of the Computer Vision and Pattern Recognition Conference, pages 20370–20382, 2025. 2

    2025

  45. [45]

    Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding

    Le Xue, Mingfei Gao, Chen Xing, Roberto Mart ´ın-Mart´ın, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, and Silvio Savarese. Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 1179–1189, 2023. 2, 3

    2023

  46. [46]

    Ulip-2: Towards scalable multimodal pre-training for 3d understanding

    Le Xue, Ning Yu, Shu Zhang, Artemis Panagopoulou, Jun- nan Li, Roberto Mart´ın-Mart´ın, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, et al. Ulip-2: Towards scalable multimodal pre-training for 3d understanding. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27091–27101, 2024. 3, 6

    2024

  47. [47]

    Po3ad: Predicting point offsets toward bet- ter 3d point cloud anomaly detection

    Jianan Ye, Weiguang Zhao, Xi Yang, Guangliang Cheng, and Kaizhu Huang. Po3ad: Predicting point offsets toward bet- ter 3d point cloud anomaly detection. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 1353–1362, 2025. 1

    2025

  48. [48]

    A unified model for multi-class anomaly detection.Advances in Neural Information Pro- cessing Systems, 35:4571–4584, 2022

    Zhiyuan You, Lei Cui, Yujun Shen, Kai Yang, Xin Lu, Yu Zheng, and Xinyi Le. A unified model for multi-class anomaly detection.Advances in Neural Information Pro- cessing Systems, 35:4571–4584, 2022. 1

    2022

  49. [49]

    Point-bert: Pre-training 3d point cloud transformers with masked point modeling

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 19313–19322, 2022. 3

    2022

  50. [50]

    Zero-shot learning in industrial scenarios: New large-scale benchmark, challenges and baseline

    Zekai Zhang, Qinghui Chen, Maomao Xiong, Shijiao Ding, Zhanzhi Su, Xinjie Yao, Yiming Sun, Cong Bai, and Jinglin Zhang. Zero-shot learning in industrial scenarios: New large-scale benchmark, challenges and baseline. InProceed- ings of the AAAI Conference on Artificial Intelligence, pages 10357–10366, 2025. 1

    2025

  51. [51]

    Pointcore: An efficient framework for unsupervised point cloud anomaly detection using joint local-global features

    Baozhu Zhao, Xiaohan Zhang, Jingfeng Guo, and Qi Liu. Pointcore: An efficient framework for unsupervised point cloud anomaly detection using joint local-global features. Neural Networks, page 108446, 2025. 7

    2025

  52. [52]

    AnomalyCLIP: Object-agnostic prompt learning for zero-shot anomaly detection

    Qihang Zhou, Guansong Pang, Yu Tian, Shibo He, and Jim- ing Chen. AnomalyCLIP: Object-agnostic prompt learning for zero-shot anomaly detection. InThe Twelfth Interna- tional Conference on Learning Representations, 2024. 2, 6, 7

    2024

  53. [53]

    Pointad: Comprehending 3d anomalies from points and pixels for zero-shot 3d anomaly detection

    Qihang Zhou, Jiangtao Yan, Shibo He, Wenchao Meng, and Jiming Chen. Pointad: Comprehending 3d anomalies from points and pixels for zero-shot 3d anomaly detection. Advances in Neural Information Processing Systems, 37: 84866–84896, 2024. 2, 6, 7

    2024

  54. [54]

    A novel ground- based cloud image segmentation method by using deep trans- fer learning.IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021

    Zecheng Zhou, Feng Zhang, Haixia Xiao, Fuchang Wang, Xin Hong, Kun Wu, and Jinglin Zhang. A novel ground- based cloud image segmentation method by using deep trans- fer learning.IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021. 1

    2021

  55. [55]

    R3d-ad: Reconstruction via diffu- sion for 3d anomaly detection

    Zheyuan Zhou, Le Wang, Naiyu Fang, Zili Wang, Lemiao Qiu, and Shuyou Zhang. R3d-ad: Reconstruction via diffu- sion for 3d anomaly detection. InEuropean conference on computer vision, pages 91–107. Springer, 2024. 1, 2, 7

    2024

  56. [56]

    Multi-granularity episodic contrastive learning for few-shot learning.Pattern Recognition, 131:108820, 2022

    Pengfei Zhu, Zhilin Zhu, Yu Wang, Jinglin Zhang, and Shuai Zhao. Multi-granularity episodic contrastive learning for few-shot learning.Pattern Recognition, 131:108820, 2022. 2

    2022

  57. [57]

    Real-iad d3: A real-world 2d/pseudo-3d/3d dataset for industrial anomaly detection

    Wenbing Zhu, Lidong Wang, Ziqing Zhou, Chengjie Wang, Yurui Pan, Ruoyi Zhang, Zhuhao Chen, Linjie Cheng, Bin- Bin Gao, Jiangning Zhang, et al. Real-iad d3: A real-world 2d/pseudo-3d/3d dataset for industrial anomaly detection. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 15214–15223, 2025. 1

    2025

  58. [58]

    Point- clip v2: Prompting clip and gpt for powerful 3d open-world learning

    Xiangyang Zhu, Renrui Zhang, Bowei He, Ziyu Guo, Ziyao Zeng, Zipeng Qin, Shanghang Zhang, and Peng Gao. Point- clip v2: Prompting clip and gpt for powerful 3d open-world learning. InProceedings of the IEEE/CVF international con- ference on computer vision, pages 2639–2650, 2023. 6, 7

    2023

This paper was first reviewed by grok-4.5 on July 14, 2026.