Anomaly Segmentation on Fishyscapes Lost & Found (val)
0.7FPR95FlowCLAS
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FlowCLAS2024.11 | 0.7 | — | — | 88.8 | — | — | — | |
| UNO2024.11 | 1.3 | — | — | 81.8 | — | — | — | |
| UNOBackbone=DINOv2-L-Rein2024.11 | 1.5 | — | — | 85.6 | — | — | — | |
| RPL2024.11 | 2.5 | — | — | 70.6 | — | — | — | |
| Dense Hybrid2026.03 | 3.9 | — | — | 47.1 | — | — | — | |
| EAM2024.11 | 4.2 | — | — | 81.5 | — | — | — | |
| Mask2Anomaly2026.03 | 4.4 | 93.6 | — | 46 | — | — | — | |
| PEBALBackbone=WideResnet382021.11 | 4.76 | 98.96 | 58.81 | — | — | — | — | |
| PEBAL2024.11 | 4.8 | — | — | 58.8 | — | — | — | |
| DenseHybrid2024.11 | 5.1 | — | — | 69.8 | — | — | — | |
| Dense Hybrid2023.07 | 6.1 | — | — | 63.8 | — | — | — | |
| RWPM2024.11 | 6.1 | — | — | 71.2 | — | — | — | |
| RbA2024.11 | 6.3 | — | — | 70.8 | — | — | — | |
| PEBALBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 6.49 | 99.09 | 59.83 | — | — | — | — | |
| PEBEL2023.07 | 6.49 | — | — | 59.83 | — | — | — | |
| PEBALBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 6.56 | 98.52 | 64.43 | — | — | — | — | |
| PEBEL2026.03 | 7.6 | — | — | 44.2 | — | — | — | |
| VL-Anomaly2026.03 | 8.4 | 96 | — | 69.5 | — | — | — | |
| Mask2Anomaly2023.07 | 9.46 | — | — | 69.41 | — | — | — | |
| Deep GamblerBackbone=WideResnet382021.11 | 10.16 | 97.82 | 31.34 | — | — | — | — | |
| MahalanobisBackbone=WideResnet382021.11 | 11.24 | 96.75 | 56.57 | — | — | — | — | |
| CosMe2021.11 | 11.65 | 98.11 | 50.22 | — | — | — | — | |
| Deep GamblerBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 12.41 | 97.19 | 39.77 | — | — | — | — | |
| Deep Gambler2023.07 | 12.41 | — | — | 39.77 | — | — | — | |
| GMMSeg-SegFormerExtra Resyn.=false, OOD Data=false, Backbone=MiT-B5 SegFormer, Generative formulation=true2022.10 | 12.55 | 97.83 | 50.03 | — | — | — | — | |
| GMMSeg-DeepLabv3+Extra Resyn.=false, OOD Data=false, Backbone=ResNet-101 DeepLabv3+, Generative formulation=true2022.10 | 13.11 | 97.34 | 43.47 | — | — | — | — | |
| MulMem2021.11 | 14.47 | 97.39 | 41.73 | — | — | — | — | |
| SML2021.11 | 14.53 | 96.88 | 36.55 | — | — | — | — | |
| SMLBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 14.53 | 96.88 | 36.55 | — | — | — | — | |
| SMLExtra Resyn.=false, OOD Data=false, Backbone=ResNet-101 DeepLabv3+2022.10 | 14.53 | 96.88 | 36.55 | — | — | — | — | |
| SML2023.07 | 14.53 | — | — | 36.55 | — | — | — | |
| GMMSeg-FCNExtra Resyn.=false, OOD Data=false, Backbone=ResNet-101 FCN, Generative formulation=true2022.10 | 16.07 | 96.28 | 32.94 | — | — | — | — | |
| AuxCon2021.11 | 18.68 | 95.79 | 19.52 | — | — | — | — | |
| SynBoost2026.03 | 18.8 | — | — | 72.6 | — | — | — | |
| SMLBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 20.09 | 96.03 | 21.71 | — | — | — | — | |
| SML2026.03 | 21.9 | — | — | 31.7 | — | — | — | |
| LDN_BINre-training=true, extra OOD data=true2021.11 | 23.97 | 95.59 | 45.71 | — | — | — | — | |
| MahalanobisExtra Resyn.=false, OOD Data=false, Backbone=ResNet-101 DeepLabv3+, Generative formulation=true2022.10 | 30.17 | 92.51 | 27.83 | — | — | — | — | |
| SynBoost2024.11 | 31 | — | — | 60.6 | — | — | — | |
| SynboostBackbone=WideResnet38, Additional learnable parameters=true, Official code results=true2021.11 | 31.02 | 96.21 | 60.58 | — | — | — | — | |
| SynBoostExtra Resyn.=true, OOD Data=true, Backbone=ResNet-101 DeepLabv3+2022.10 | 31.02 | 96.21 | 60.58 | — | — | — | — | |
| EnergyBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 32.26 | 93.5 | 25.79 | — | — | — | — | |
| Energy2023.07 | 32.26 | — | — | 25.79 | — | — | — | |
| MSPBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 32.55 | 89.26 | 11.84 | — | — | — | — | |
| SMLBackbone=WideResnet38, Official code results=true2021.11 | 33.49 | 94.97 | 22.74 | — | — | — | — | |
| SynboostBackbone=Resnet101, additional learnable parameters=true, official code with Resnet101 backbone=true2021.11 | 34.47 | 94.89 | 40.99 | — | — | — | — | |
| SynBoost2023.07 | 34.47 | — | — | 40.99 | — | — | — | |
| Max Entropy2026.03 | 35.1 | — | — | 29.9 | — | — | — | |
| Meta-OoDBackbone=WideResnet382021.11 | 37.69 | 93.06 | 41.31 | — | — | — | — | |
| EnergyBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 37.71 | 93.45 | 14.29 | — | — | — | — | |
| MaxLogit2021.11 | 38.13 | 92 | 18.77 | — | — | — | — | |
| Max LogitBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 38.13 | 92 | 18.77 | — | — | — | — | |
| Max LogitBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 38.15 | 93.14 | 12.78 | — | — | — | — | |
| EntropyBackbone=WideResnet382021.11 | 40.34 | 90.82 | 10.36 | — | — | — | — | |
| Entropy2023.07 | 40.34 | — | — | 10.36 | — | — | — | |
| MSPBackbone=WideResnet382021.11 | 40.59 | 89.29 | 4.59 | — | — | — | — | |
| Max Softmax2023.07 | 40.59 | — | — | 4.59 | — | — | — | |
| EnergyBackbone=WideResnet382021.11 | 41.78 | 93.72 | 16.05 | — | — | — | — | |
| Max LogitBackbone=WideResnet382021.11 | 42.21 | 93.41 | 14.59 | — | — | — | — | |
| Max Logit2023.07 | 42.21 | — | — | 14.59 | — | — | — | |
| MSP2026.03 | 44.8 | — | — | 1.3 | — | — | — | |
| Entropy2026.03 | 44.8 | — | — | 2.9 | — | — | — | |
| EntropyBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 44.85 | 88.32 | 13.91 | — | — | — | — | |
| EntropyExtra Resyn.=false, OOD Data=false, Backbone=ResNet-101 DeepLabv3+2022.10 | 44.85 | 88.32 | 13.91 | — | — | — | — | |
| MSP2021.11 | 45.63 | 86.99 | 6.02 | — | — | — | — | |
| MSPBackbone=Resnet101, additional learnable parameters=false, official code with Resnet101 backbone=false2021.11 | 45.63 | 86.99 | 6.02 | — | — | — | — | |
| MSPExtra Resyn.=false, OOD Data=false, Backbone=ResNet-101 DeepLabv3+2022.10 | 45.63 | 86.99 | 6.02 | — | — | — | — | |
| SynthCP2021.11 | 45.95 | 88.34 | 6.54 | — | — | — | — | |
| SynthCPBackbone=Resnet101, additional learnable parameters=true, official code with Resnet101 backbone=true2021.11 | 45.95 | 88.34 | 6.54 | — | — | — | — | |
| SynthCPExtra Resyn.=true, OOD Data=false, Backbone=ResNet-101 DeepLabv3+2022.10 | 45.95 | 88.34 | 6.54 | — | — | — | — | |
| SynthCP2023.07 | 45.95 | — | — | 6.54 | — | — | — | |
| EntropyBackbone=WideResNet38, Split=cv0 standard, Base Model=DeepLabv3+ (Nvidia)2021.11 | 47.81 | 89.01 | 8.79 | — | — | — | — | |
| ODIN2026.03 | 81.1 | 86.3 | — | 1.4 | — | — | — | |
| FastFlowBackbone=DINOv2-L-Rein2024.11 | 94.4 | — | — | 8.9 | — | — | — | |
| DenseHybridthresholding=optimal2023.11 | — | — | — | — | 2.41 | 15.35 | 3.83 | |
| PEBALthresholding=optimal2023.11 | — | — | — | — | 1.49 | 6.35 | 2.57 | |
| RPL+CoroCLthresholding=optimal2023.11 | — | — | — | — | 2.55 | 15.75 | 3.91 | |
| S2Mprompt generator=Faster R-CNN (ResNet-50), segmentation model=SAM (ViT-B), base anomaly score=RPL2023.11 | — | — | — | — | 29.3 | 30.46 | 35.31 | |
| Synboostthresholding=optimal2023.11 | — | — | — | — | 7.37 | 18.39 | 10.85 |