Anomaly Detection on MVTec AD (image-level AUROC)
99.72AUROC (Image-level)ReConPatch
Evaluation Results
| Method | Links | |
|---|---|---|
| ReConPatchBackbone=WRN-101 & RNext-101 & DenseN-201, Image size=480×4802023.05 | 99.72 | |
| CPR2024.07 | 99.7 | |
| ReConPatchBackbone=WRN-101 & RNext-101 & DenseN-201, Image size=320×3202023.05 | 99.67 | |
| PNI (w/ refine)Backbone=WRN-101 & RNext-101 & DenseN-201, Image size=480×4802023.05 | 99.63 | |
| PatchCoreBackbone=WRN-101 & RNext-101 & DenseN-201, Image size=320×3202023.05 | 99.6 | |
| ADClickSupervision=weak2024.07 | 99.6 | |
| MemKD2024.08 | 99.6 | |
| PRN2024.07 | 99.4 | |
| RD++2024.08 | 99.4 | |
| DMDD2024.08 | 99.4 | |
| BGAD2024.07 | 99.3 | |
| THFRRequires Training=Yes, Network=CNN, Structural Capability=Yes, Logical Capability=Yes, FPS=7.692024.06 | 99.2 | |
| PatchcoreRequires Training=Yes, Network=CNN, Structural Capability=Yes, Logical Capability=No, FLOPs (Gb)=11.4, FPS=25.12024.06 | 99.1 | |
| DeSTSeg2024.08 | 98.6 | |
| GLCFRequires Training=Yes, Network=Transformer, Structural Capability=Yes, Logical Capability=Yes, FLOPs (Gb)=52.6, FPS=11.22024.06 | 98.6 | |
| RD2024.08 | 98.5 | |
| SAM-LADRequires Training=No, Network=Transformer, Structural Capability=Yes, Logical Capability=Yes, FLOPs (Gb)=54.7, FPS=8.92024.06 | 98.4 | |
| CRAD2026.02 | 98.2 | |
| DPDLNumber of training anomaly examples=102025.02 | 97.7 | |
| HLGFA2026.02 | 97.5 | |
| AHLNumber of training anomaly examples=102025.02 | 97 | |
| DRA2024.07 | 95.9 | |
| DRANumber of training anomaly examples=102025.02 | 95.9 | |
| STPM2024.08 | 95.5 | |
| PadimRequires Training=Yes, Network=CNN, Structural Capability=Yes, Logical Capability=No, FPS=4.62024.06 | 95.5 | |
| Pull & Push2024.08 | 94.8 | |
| DevNet2024.07 | 94.5 | |
| DevNetNumber of training anomaly examples=102025.02 | 94.5 | |
| NGAL2026.02 | 94.5 | |
| FLOSNumber of training anomaly examples=102025.02 | 93.9 | |
| DPDLNumber of training anomaly examples=12025.02 | 92.7 | |
| SAOENumber of training anomaly examples=102025.02 | 92.6 | |
| RD4AD2026.02 | 92.1 | |
| AnomalyCLIP2026.02 | 91.6 | |
| MLEPNumber of training anomaly examples=102025.02 | 90.7 | |
| AHLNumber of training anomaly examples=12025.02 | 90.1 | |
| DRANumber of training anomaly examples=12025.02 | 88.3 | |
| US2024.08 | 87.7 | |
| SPADERequires Training=Yes, Network=CNN, Structural Capability=Yes, Logical Capability=No, FPS=0.92024.06 | 85.5 | |
| SAOENumber of training anomaly examples=12025.02 | 83.4 | |
| DevNetNumber of training anomaly examples=12025.02 | 78 | |
| FLOSNumber of training anomaly examples=12025.02 | 75.5 | |
| MLEPNumber of training anomaly examples=12025.02 | 74.4 | |
| AERequires Training=No, Network=CNN, Structural Capability=Yes, Logical Capability=No, FLOPs (Gb)=5.0, FPS=251.12024.06 | 71 | |
| f-AnoGANRequires Training=No, Network=CNN, Structural Capability=Yes, Logical Capability=No, FLOPs (Gb)=7.7, FPS=133.42024.06 | 65.8 |