Efficient3D-AD: Token-Efficient and View-Aware Zero-shot 3D Multimodal Anomaly Detection
Abstract
Zero-shot 3D multimodal anomaly detection is a crucial yet underexplored problem. Existing approaches typically render point clouds into multiple fixed-view images and extract features using pretrained networks such as Vision Transformers (ViTs). However, rendering many views increases computational cost, and redundant views can introduce uninformative or misleading features that reduce detection accuracy. To address these challenges, we propose an efficient zero-shot 3D anomaly detection framework with two key components. First, a task-driven adaptive multi-view rendering module identifies the most discriminative viewpoints, enhancing the ability of representations to capture anomalies. Second, a dynamic token pruning mechanism removes redundant tokens while focusing on anomalous regions. Experiments demonstrate that our method significantly lowers computational cost while achieving performance comparable to the baseline models.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
PointCLIP: Point Cloud Understanding by CLIP
Zhang, R., Guo, Z., Zhang, W., Li, K., Miao, X., Cui, B., Qiao, Y., Gao, P., & Li, H. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00836 · Cited by 483 · see citing works
The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization
Bergmann, P., Jin, X., Sattlegger, D., & Steger, C. T. (2022) · International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP) · doi:10.5220/0010865000003124 · Cited by 234 · see citing works
Complementary pseudo multimodal feature for point cloud anomaly detection
Cao, Y., Xu, X., & Shen, W. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110761 · Cited by 119 · see citing works
3D-MMFN: Multi-level multimodal fusion network for 3D industrial image anomaly detection
Asad, M., Azeem, W., Malik, A. A., Jiang, H., Ali, A., Yang, J., & Liu, W. (2025) · Advanced Engineering Informatics · doi:10.1016/j.aei.2025.103284 · Cited by 48 · see citing works
R3D-AD: Reconstruction via Diffusion for 3D Anomaly Detection
Zhou, Z., Wang, L., Fang, N., Wang, Z., Qiu, L., & Zhang, S. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72764-1_6 · Cited by 35 · see citing works
3D vision-based anomaly detection in manufacturing: A survey
Du, J., Tao, C., Cao, X., & Tsung, F. (2025) · Frontiers of Engineering Management · doi:10.1007/s42524-025-4189-9 · Cited by 30 · see citing works
2M3DF: Advancing 3D Industrial Defect Detection With Multi-Perspective Multimodal Fusion Network
Asad, M., Azeem, W., Jiang, H., Mustafa, H. T., Yang, J., & Liu, W. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3536475 · Cited by 24 · see citing works
Dynamic Tuning Towards Parameter and Inference Efficiency for ViT Adaptation
Zhao, W., Tang, J., Han, Y., Song, Y., Wang, K., Huang, G., Wang, F., & You, Y. (2024) · neural information processing systems · doi:10.52202/079017-3643 · Cited by 17 · see citing works
PointAD: Comprehending 3D Anomalies from Points and Pixels for Zero-shot 3D Anomaly Detection
Zhou, Q., Yan, J., He, S., Meng, W., & Chen, J. (2024) · neural information processing systems · doi:10.52202/079017-2695 · Cited by 16 · see citing works
Self-citationToward Zero-Shot Point Cloud Anomaly Detection: A Multiview Projection Framework
Cheng, Y., Cao, Y., Xie, G., Lu, Z., & Shen, W. (2025) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2025.3648581 · Cited by 8 · see citing works
Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction
Zhao, S., Wang, Z., Juefei-Xu, F., Xia, X., Liu, M., Wang, X., Liang, M., Zhang, N., Metaxas, D., & Yu, L. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.02780 · Cited by 8 · see citing works
Zero-shot 3D anomaly detection via online voter mechanism
Zheng, W., Ke, X., & Guo, W. (2025) · Neural Networks · doi:10.1016/j.neunet.2025.107398 · Cited by 5 · see citing works
Multi-View Reconstruction with Global Context for 3D Anomaly Detection *
Sun, Y., Cheng, Y., Cao, Y., Zhang, Y., & Shen, W. (2025) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc58881.2025.11343209 · Cited by 5 · see citing works
Abhisri, M., Sasidhar, M., Subramanian, A., Sreeraghavan, M., Sivakumar, S., & Prabakaran, N. (2024) · International Conference on Recent Advances in Science and Engineering Technology (ICRASET) · doi:10.1109/icraset63057.2024.10895730 · Cited by 4 · see citing works
CATP: Cross-Attention Token Pruning for Accuracy Preserved Multimodal Model Inference
Liao, R., Zhao, C., Jin, L., Feng, W., Lyu, Y., Chen, B., & Yang, H. (2025) · IEEE Conference on Artificial Intelligence (CAI) · doi:10.1109/cai64502.2025.00191 · Cited by 3 · see citing works
Efficient slice anomaly detection network for 3D brain MRI Volume
Zhang, Z., & Mohsenzadeh, Y. (2025) · PLOS Digital Health · doi:10.1371/journal.pdig.0000874 · Cited by 2 · see citing works
Flagged first, then by citations
Cited by
Still being checked.
- Authors
- Chen, J., Gu, L., Chen, X., & Li, Z.
- Affiliations
- Beijing University of Posts and Telecommunications (CN)
- Objecteye.ltd
- Beihang University (CN)
- ORCID
- Venue
- ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
- Publisher
- IEEE
- Published
- 2026-05-03
- Pages
- 11247-11251
- ISSN
- 1520-6149 (ISSN-L)
- OpenAlex
- W7155069050
- Cited by
- 0
- References
- 16
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE International Conference on Acoustics Speech and Signal Processing (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Anomaly Detection Techniques and Applications
- Physical Sciences › Engineering › Computational Mechanics › 3D Shape Modeling and Analysis
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- zero-shot anomaly detection, point clouds, token pruning, computational complexity reduction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
