Pith. sign in

REVIEW 11 cited by

The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2112.09045 v1 pith:QK73LTEN submitted 2021-12-16 cs.CV

The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization

classification cs.CV
keywords anomalydatasetdefectsdetectionobjectanomaly-freecategoriesdata
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

We introduce the first comprehensive 3D dataset for the task of unsupervised anomaly detection and localization. It is inspired by real-world visual inspection scenarios in which a model has to detect various types of defects on manufactured products, even if it is trained only on anomaly-free data. There are defects that manifest themselves as anomalies in the geometric structure of an object. These cause significant deviations in a 3D representation of the data. We employed a high-resolution industrial 3D sensor to acquire depth scans of 10 different object categories. For all object categories, we present a training and validation set, each of which solely consists of scans of anomaly-free samples. The corresponding test sets contain samples showing various defects such as scratches, dents, holes, contaminations, or deformations. Precise ground-truth annotations are provided for every anomalous test sample. An initial benchmark of 3D anomaly detection methods on our dataset indicates a considerable room for improvement.

discussion (0)

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

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. CMDS-AD: Cross-Modal Dual-Stream Decoupling for Few-Shot Anomaly Detection

    cs.CV 2026-06 unverdicted novelty 7.0

    CMDS-AD decouples low-frequency normals via diffusion estimator from high-frequency defects in real RGB data using dual streams, a hierarchical mapper, and multiplicative scoring to achieve SOTA 1-shot anomaly detecti...

  2. Real-IAD MVN: A Multi-View Normal Vector Dataset and Benchmark for High-Fidelity Industrial Anomaly Detection

    cs.CV 2026-05 unverdicted novelty 7.0

    Real-IAD-MVN supplies multi-view normal vector data and a reconstruction baseline that outperforms prior multimodal methods for geometric industrial anomaly detection.

  3. Align3D-AD: Cross-Modal Feature Alignment and Dual-Prompt Learning for Zero-shot 3D Anomaly Detection

    cs.CV 2026-05 unverdicted novelty 7.0

    Align3D-AD improves zero-shot 3D anomaly detection by cross-modal feature alignment from RGB guidance and dual-prompt contrastive alignment to capture complementary semantics.

  4. MMR-AD: A Large-Scale Multimodal Dataset for Benchmarking General Anomaly Detection with Multimodal Large Language Models

    cs.CV 2026-04 unverdicted novelty 7.0

    MMR-AD is a new benchmark dataset showing that current generalist MLLMs lag industrial needs for anomaly detection, with Anomaly-R1 delivering better results through reasoning and RL.

  5. Anomaly Factory 3D: A Modular Framework for Diverse Pseudo-Anomaly Synthesis in Unsupervised 3D Anomaly Detection

    cs.CV 2026-06 unverdicted novelty 6.0

    AF3AD is a modular synthesis framework using center-conditioned parametric deformations in local PCA frames to create diverse pseudo-anomalies, improving unsupervised 3D anomaly detection on AnomalyShapeNet and Real3D-AD.

  6. Learning Topology-Aware Representations via Test-Time Adaptation for Anomaly Segmentation

    cs.CV 2026-06 unverdicted novelty 6.0

    TopoTTA integrates persistent homology into test-time adaptation to derive topological pseudo-labels from anomaly maps, improving segmentation by an average 15% F1 on six benchmarks while generalizing across 2D and 3D data.

  7. CoGeoAD: Hierarchical Color-Geometric Fusion with Multi-View Attention for Zero-Shot 3D Anomaly Detection

    cs.CV 2026-06 unverdicted novelty 6.0

    CoGeoAD fuses color and geometric features via Data-Driven Multi-View Attention and Multi-Stage Color-Geometric Fusion in a CLIP backbone to reach SOTA zero-shot 3D anomaly detection on MVTec3D-AD and Eyecandies.

  8. VT-3DAD: Cross-Category 3D Anomaly Detection via Visual-Text Normal Space Alignment

    cs.CV 2026-06 unverdicted novelty 6.0

    VT-3DAD fuses visual deviation from few-shot references and semantic deviation from textual normal space to achieve SOTA cross-category 3D anomaly detection on ShapeNetPart.

  9. Advancing Pre-trained Teacher: Towards Robust Feature Discrepancy for Anomaly Detection

    cs.CV 2024-05 unverdicted novelty 6.0

    AAND is a two-stage anomaly detection method that advances a pre-trained teacher via residual anomaly amplification and applies hard knowledge distillation in reverse distillation to achieve SOTA results on MVTecAD, V...

  10. Parameter Efficient Multi-Class Intelligent Scheduling for Multimodal Online Distributed Industrial Anomaly Detection

    cs.LG 2026-05 unverdicted novelty 5.0

    Proposes MODIAD framework with MIS scheduling solved via SMG algorithm and REC-LoRA adaptation for efficient multimodal online distributed industrial anomaly detection, reporting superior performance on MVTec 3D-AD an...

  11. ZSG-IAD: A Multimodal Framework for Zero-Shot Grounded Industrial Anomaly Detection

    cs.CV 2026-04 unverdicted novelty 5.0

    ZSG-IAD is a zero-shot multimodal system that uses language-guided two-hop grounding and rule-based reinforcement learning to produce anomaly masks and explainable reports from industrial sensor data.