Awesome-3D-AnomalyDetection

Anomaly detection on 3D point cloud data has been a rising topic recently and has wide-range applications in real-world applications, especially in industrial scenarios. This repository summarizes the state-of-the-art methods and the related papers. The literature are listed in alphabet order for each section.

Dataset

[MVTec 3D-AD] | The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization | VISAPP' 22 | [pdf] [website]

[Eyecandies] | The Eyecandies Dataset for Unsupervised Multimodal Anomaly Detection and Localization | ACCV' 22 | [pdf] [website]

[Real3D-AD] | Real3D-AD: A Dataset of Point Cloud Anomaly Detection | NeurIPS' 23 | [pdf] [website]

Unsupervised 3D Anomaly Detection (and Localization) on MVTec 3D-AD Dataset

Performance Comparison on MvTEC 3D

Time	Method	Image-level AUROC (%)	Pixel-level AUPRO (%)
2021	3D-Only-Voxel-VM	69.9	49.2
2021	3D+RGB-Voxel-VM	51.7	63.9
2022	BTF	86.5	96.4
2022	3D-ST	-	83.3
2022	RGB+PaDim	76.4	93.0
2022	RGB+PatchCore	81.1	91.0
2022	AST	93.7	-
2022	PEFM	-	94.2
2023	CDO	-	93.8
2023	M3DM	94.5	96.4
2023	CPMF	95.2	92.9
2023	Shape-Guided	94.7	97.6

• [3D-Only-Voxel-VM,3D+RGB-Voxel-VM] The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization | VISAPP' 22 | [pdf] [website]
• [BTF] Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection | arXiv' 22 | [pdf] [code]
• [3D-ST] Anomaly Detection in 3D Point Clouds Using Deep Geometric Descriptors | WACV' 23 | [pdf]
• [RGB+PaDim,RGB+PatchCore] Benchmarking Unsupervised Anomaly Detection and Localization | arXiv' 22 | [pdf]
• [AST] Asymmetric Student-Teacher Networks for Industrial Anomaly Detection | WACV' 23 | [pdf] [code]
• [PEFM] Position Encoding Enhanced Feature Mapping for Image Anomaly Detection | IEEE 18th International Conference on Automation Science and Engineering (CASE) | [pdf]
• [CDO] Collaborative Discrepancy Optimization for Reliable Image Anomaly Localization | IEEE Transactions on Industrial Informatics | [pdf] [code]
• [M3DM] Multimodal Industrial Anomaly Detection via Hybrid Fusion | CVPR' 23 | [pdf] [code]
• [CPMF] Complementary Pseudo Multimodal Feature for Point Cloud Anomaly Detection | Pattern Recognition' 23 (under review) | [pdf] [code]
• [Shape-Guided] Complementary Pseudo Multimodal Feature for Point Cloud Anomaly Detection | ICML' 23 | [pdf] [code]

3D Anomaly Detection on Other Datasets

• Anomaly detection in 3D space for autonomous driving | KIT Dissertation Paper | [PDF]
• Toward Unsupervised 3d Point Cloud Anomaly Detection Using Variational Autoencoder | ICIP' 21| [PDF]