Weakly Supervised Object Localization on ImageNet-1K (val)
65.2Top-1 Loc AccGenPromp
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GenPrompLoc Back.=Stable Diffusion, Cls Back.=EfficientNet-B7, Prompt ensemble=true2023.07 | 65.2 | 73.4 | 75 | — | |
| GenPrompLoc Back.=Stable Diffusion, Cls Back.=EfficientNet-B7, Prompt ensemble=false2023.07 | 65.1 | 73.3 | 74.9 | — | |
| C2AMLoc Back.=DenseNet161, Cls Back.=EfficientNet-B72023.07 | 59.6 | 67.1 | 68.5 | — | |
| PSOLLoc Back.=DenseNet161, Cls Back.=EfficientNet-B72023.07 | 58 | 65 | 66.3 | — | |
| BASLoc Back.=ResNet502023.07 | 57.2 | 67.4 | 71.8 | — | |
| LCTRALoc Back.=Deit-S2023.07 | 56.1 | 65.8 | 68.7 | — | |
| SCMLoc Back.=Deit-S2023.07 | 56.1 | 66.4 | 68.8 | — | |
| CREAMLoc Back.=InceptionV32023.07 | 56.1 | 66.2 | 69 | — | |
| TS-CAMLoc Back.=Deit-S2023.07 | 53.4 | 64.3 | 67.6 | — | |
| CAMLoc Back.=VGG162023.07 | 42.8 | 54.9 | 59 | — | |
| Co-MixupBackbone=ResNet-50, Training Epochs=100, Training speed increment=3/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 55.32 | |
| CutMixBackbone=ResNet-50, Training Epochs=100, Training speed increment=1/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 54.91 | |
| InputBackbone=ResNet-50, Training Epochs=100, Training speed increment=1/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 55.07 | |
| ManifoldBackbone=ResNet-50, Training Epochs=100, Training speed increment=1/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 54.86 | |
| PuzzleMixBackbone=ResNet-50, Training Epochs=100, Training speed increment=2.8/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 55.22 | |
| R-MixBackbone=ResNet-50, Training Epochs=100, Training speed increment=2/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 55.58 | |
| VanillaBackbone=ResNet-50, Training Epochs=100, Training speed increment=1/1, Evaluation Protocol=Class Activation Map (Zhou et al. 2015)2022.12 | — | — | — | 54.36 |