Anomaly Detection on CIFAR-100
98.97AUROCCLIP-ViT-L/14 (LAION-2B)
Evaluation results
| Method | Links | |||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CLIP-ViT-L/14 (LAION-2B)Transform=gLocal2023.06 | 98.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14 (LAION-2B)Transform=original2023.06 | 98.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14 (LAION-400M)Transform=original2023.06 | 98.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14 (LAION-400M)Transform=gLocal2023.06 | 98.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeneralADModel Size=large, Backbone=DINOv22024.07 | 98.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FITYMI2024.07 | 98.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeneralADModel Size=base, Backbone=DINOv22024.07 | 98 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Transformalysetting=unimodal2021.12 | 97.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KDADBackbone=DINOv2-B2024.07 | 97.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransformalyLearning Paradigm=Pretrained, Setting=Unimodal2021.12 | 97.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Transformaly2024.07 | 97.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDA-OESupervision Setting=OE2020.10 | 97.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14 (WIT)Transform=gLocal2023.06 | 97.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeneralADModel Size=small, Backbone=DINOv22024.07 | 96.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CAPadaptation=true2021.12 | 96.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSADLearning Paradigm=Pretrained, Setting=Unimodal2021.12 | 96.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSADsetting=unimodal2021.12 | 96.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSAD2024.07 | 96.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CAPadaptation=false2021.12 | 96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN22021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDA2021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN2Learning Paradigm=Pretrained, Setting=Unimodal2021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDALearning Paradigm=Pretrained, Setting=Unimodal2021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN2setting=unimodal2021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDAsetting=unimodal2021.12 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDA2024.07 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN2Supervision Setting=Pretrained2020.10 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDASupervision Setting=Pretrained2020.10 | 94.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-RN50 (WIT)Transform=gLocal2023.06 | 92.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14 (WIT)Transform=original2023.06 | 91.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIP-RN50 (WIT)Transform=original2023.06 | 90.83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Outlier Exposure2021.12 | 90.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSILearning Paradigm=Self-Supervised, Setting=Unimodal2021.12 | 89.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSIsetting=unimodal2021.12 | 89.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PIXMIX2021.12 | 89.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDAD2024.07 | 86.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AugMix2021.12 | 84.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Recontrast2024.07 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Auto Augment2021.12 | 80.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MHRot2021.12 | 80.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MHRotLearning Paradigm=Self-Supervised, Setting=Unimodal2021.12 | 80.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MHRotSupervision Setting=Self-Supervised2020.10 | 80.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Transformalysetting=multimodal2021.12 | 79.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ARNetProtocol=One-class training protocol2019.11 | 78.8 | — | 77.5 | 70 | 62.4 | 76.2 | 77.7 | 64 | 86.9 | 65.6 | 82.7 | 90.2 | 85.9 | 83.5 | 84.6 | 67.6 | 84.2 | 74.1 | 80.3 | 91 | 85.3 | 85.4 | 8.82 | |
| Geometric TransformationsNormal Class Index (G)=avg2018.05 | 78.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeoTransProtocol=One-class training protocol2019.11 | 78.7 | — | 74.7 | 68.5 | 74 | 81 | 78.4 | 59.1 | 81.8 | 65 | 85.5 | 90.6 | 87.6 | 83.9 | 83.2 | 58 | 92.1 | 68.3 | 73.5 | 93.8 | 90.7 | 85 | 10.76 | |
| Baseline2021.12 | 77.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CutMix2021.12 | 74.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cutout2021.12 | 74.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mixup2021.12 | 71.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN2setting=multimodal2021.12 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleNet2024.07 | 69.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MSADsetting=multimodal2021.12 | 67.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Deep-SVDD2021.12 | 67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepSVDDLearning Paradigm=Unsupervised, Setting=Unimodal2021.12 | 67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepSVDDSupervision Setting=Self-Supervised2020.10 | 67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Patchcore2024.07 | 64.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC-SVM (CAE)Normal Class Index (G)=avg, Hyperparameter Optimization=Hindsight knowledge2018.05 | 63.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC-SVM (RAW)Normal Class Index (G)=avg, Hyperparameter Optimization=Hindsight knowledge2018.05 | 62.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC-SVMLearning Paradigm=Unsupervised, Setting=Unimodal2021.12 | 62.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC-SVMSupervision Setting=Self-Supervised2020.10 | 62.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PANDAsetting=multimodal2021.12 | 62.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ALOCCProtocol=One-class training protocol2019.11 | 60.9 | — | 52.7 | 56.6 | 61.2 | 61.1 | 66.7 | 50.6 | 63.9 | 66.2 | 50.9 | 73.4 | 71.1 | 56.9 | 63.6 | 56 | 57.9 | 58.2 | 57 | 73.5 | 61 | 58.8 | 6.7 | |
| DSEBMNormal Class Index (G)=avg2018.05 | 58.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSEBMProtocol=One-class training protocol2019.11 | 58.8 | — | 64 | 47.9 | 53.7 | 48.4 | 59.7 | 46.6 | 51.7 | 54.8 | 66.7 | 71.2 | 78.3 | 62.7 | 66.8 | 52.6 | 44 | 56.8 | 63.1 | 73 | 57.7 | 55.5 | 9.36 | |
| GANomalyProtocol=One-class training protocol2019.11 | 56.5 | — | 57.9 | 51.9 | 36 | 46.5 | 46.6 | 42.9 | 53.7 | 59.4 | 63.7 | 68 | 75.6 | 57.6 | 58.7 | 59.9 | 43.9 | 59.9 | 64.4 | 71.8 | 54.9 | 56.8 | 9.94 | |
| AD-GANNormal Class Index (G)=avg2018.05 | 54.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ADGANProtocol=One-class training protocol2019.11 | 54.7 | — | 63.1 | 54.9 | 41.3 | 50 | 40.6 | 42.8 | 51.1 | 55.4 | 59.2 | 62.7 | 79.8 | 53.7 | 58.9 | 57.4 | 39.4 | 55.6 | 63.3 | 66.7 | 44.3 | 53 | 10.08 | |
| AEProtocol=One-class training protocol2019.11 | 52.4 | — | 66.7 | 55.4 | 41.4 | 49.2 | 44.9 | 40.6 | 50.2 | 48.1 | 66.1 | 63 | 52.7 | 62.1 | 59.6 | 49.8 | 48.1 | 56.4 | 57.6 | 47.2 | 47.1 | 41.5 | 8.11 | |
| Deep SVDDsetting=multimodal2021.12 | 50.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAGMMNormal Class Index (G)=avg2018.05 | 50.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAGMMProtocol=One-class training protocol2019.11 | 50.5 | — | 43.4 | 49.5 | 66.1 | 52.6 | 56.9 | 52.4 | 55 | 52.8 | 53.2 | 42.5 | 52.7 | 46.4 | 42.7 | 45.4 | 57.2 | 48.8 | 54.4 | 36.4 | 52.4 | 50.3 | 6.55 |