Out-of-distribution image detection on SVHN (test)
0.1FPR95MSP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MSPOOD Detector=Maximum Softmax Probability, Backbone=Wide Residual Network, Outlier Exposure=true2018.12 | 0.1 | 100 | 99.9 | |
| MSP with OECCBackbone=WRN-16-4, Fine-tuning method=OECC, Detector type=Maximum Softmax Probability (MSP)2019.06 | 3 | 99.99 | 99.55 | |
| MSPOOD Detector=Maximum Softmax Probability, Backbone=Wide Residual Network, Outlier Exposure=false2018.12 | 6.3 | 98 | 91.1 | |
| MSP with Outlier ExposureBackbone=WRN-16-4, Fine-tuning method=Outlier Exposure (OE), Detector type=Maximum Softmax Probability (MSP)2019.06 | 10 | 99.98 | 99.83 |