Metric: Detection AUROC (higher is better)
| # | Model↕ | Detection AUROC▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | AST | 93.7 | Yes | Asymmetric Student-Teacher Networks for Industri... | 2022-10-14 | Code |
| 2 | TransFusion | 0.957 | No | TransFusion -- A Transparency-Based Diffusion Mo... | 2023-11-16 | Code |
| 3 | MC4AD | 0.954 | Yes | Examining the Source of Defects from a Mechanica... | 2025-05-09 | Code |
| 4 | CPMF (2D+3D) | 0.9515 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 5 | CPMF (2D+3D) | 0.9515 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 6 | GLFM | 0.944 | No | - | - | Code |
| 7 | 3DSR (3D) | 0.922 | No | Cheating Depth: Enhancing 3D Surface Anomaly Det... | 2023-11-02 | Code |
| 8 | Shape-Guided (only SDF) | 0.916 | No | - | - | Code |
| 9 | CPMF (2D) | 0.8918 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 10 | CPMF (2D) | 0.8918 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 11 | CPMF (3D) | 0.8304 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 12 | CPMF (3D) | 0.8304 | Yes | Complementary Pseudo Multimodal Feature for Poin... | 2023-03-23 | Code |
| 13 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (FPFH) | 0.782 | No | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 14 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (SIFT) | 0.727 | No | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 15 | Voxel AE | 0.699 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |
| 16 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (NSA) | 0.696 | No | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 17 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (Depth iNet) | 0.675 | Yes | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 18 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (RaW) | 0.573 | No | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 19 | Voxel VM | 0.571 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |
| 20 | Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (HoG) | 0.559 | No | Back to the Feature: Classical 3D Features are (... | 2022-03-10 | Code |
| 21 | Depth VM | 0.546 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |
| 22 | Depth AE | 0.546 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |
| 23 | Voxel GAN | 0.537 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |
| 24 | Depth GAN | 0.523 | No | The MVTec 3D-AD Dataset for Unsupervised 3D Anom... | 2021-12-16 | Code |