Efficiency Analysis on ImageNet-100 (train)
0.13Storage Usage (GB)Collab
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CollabHardware=Two NVIDIA V100 (32GB) GPUs2021.03 | 0.13 | 5.58 | 5.23 | 10.81 | |
| DRSHardware=Two NVIDIA V100 (32GB) GPUs, Components=Discriminator network2021.03 | 0.13 | 1.47 | 0.75 | 2.22 | |
| DDLSHardware=Two NVIDIA V100 (32GB) GPUs2021.03 | 0.13 | 0 | 218.05 | 218.05 | |
| cDR-RSHardware=Two NVIDIA V100 (32GB) GPUs, Backbone=ResNet-34, Density Ratio Model=MLP-5, Training Epochs=2002021.03 | 0.75 | 65.69 | 1.18 | 66.87 | |
| DRE-F-SP+RSHardware=Two NVIDIA V100 (32GB) GPUs, Backbone=ResNet-34, Density Ratio Model=100 MLP-5 models, Training Epochs=4002021.03 | 38.94 | 84.41 | 2.43 | 86.84 |