Low-shot Image Classification on Places 205 (k-shot evaluation)
22.2Accuracy (k=1)MoPro
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MoProPretrain dataset=WebVision-V2.0, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 22.2 | 29.2 | 35.6 | 40.9 | 44.4 | |
| CEPretrain dataset=WebVision-V2.0, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 21.8 | 28.6 | 35.1 | 40 | 43.6 | |
| MoProPretrain dataset=WebVision-V1.0, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 16.9 | 23.2 | 29.2 | 34.5 | 38.7 | |
| CEPretrain dataset=WebVision-V1.0, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 15.1 | 21.5 | 27.2 | 32.1 | 36.4 | |
| MoProPretrain dataset=WebVision-V0.5, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 15 | 21.2 | 26.6 | 31.8 | 36 | |
| CE (Sup.)Pretrain dataset=ImageNet, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 14.9 | 21 | 26.9 | 32.1 | 36 | |
| CEPretrain dataset=WebVision-V0.5, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 13.5 | 19.3 | 24.7 | 29.5 | 33.8 | |
| PCL v2*Pretrain dataset=ImageNet, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 12.5 | 17.5 | 23.2 | 28.1 | 32.3 | |
| MoCo v2*Pretrain dataset=ImageNet, Classifier=Linear SVM, Feature type=Fixed representation2020.09 | 11.9 | 17 | 22.6 | 28.1 | 32.4 |