Object Localization on ImageNet-1k (val)
65.2Top-1 Loc AccGenPrompt
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GenPromptLoc Back.=Stable Diffusion, Cls Back.=EfficientNet-B7, Prompt ensemble strategy=true2023.07 | 65.2 | 73.4 | 75 | |
| GenPromptLoc Back.=Stable Diffusion, Cls Back.=EfficientNet-B72023.07 | 65.1 | 73.3 | 74.9 | |
| C2AMLoc Back.=DenseNet161, Cls Back.=EfficientNet-B72023.07 | 59.6 | 67.1 | 68.5 | |
| SPOLLoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 59.1 | 67.2 | 69 | |
| BASLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 58.5 | 69 | 71.9 | |
| PSOL-DenseNet161-SepClassification model=EfficientNet-B72020.02 | 58 | 65.02 | — | |
| PSOLLoc Back.=DenseNet161, Cls Back.=EfficientNet-B72023.07 | 58 | 65 | 66.3 | |
| TAFormerLoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 57.5 | 69.9 | 75.5 | |
| PSOL-InceptionV3-SepClassification model=EfficientNet-B72020.02 | 57.25 | 64.04 | — | |
| BASLoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 57.2 | 67.4 | 71.8 | |
| PSOL-DenseNet161-SepClassification model=DPN1312020.02 | 56.59 | 64.63 | — | |
| SPGClassification model=DPN-ensemble2020.02 | 56.17 | 63.22 | — | |
| CREAMLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 56.1 | 66.2 | 69 | |
| LCTRLoc Back.=Deit-S, Cls Back.=Deit-S2023.07 | 56.1 | 65.8 | 68.7 | |
| SCMELoc Back.=Deit-S, Cls Back.=Deit-S2023.07 | 56.1 | 66.4 | 68.8 | |
| TAFormerLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 56 | 66.5 | 69.8 | |
| DALoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 55.8 | — | 70.3 | |
| PSOL-InceptionV3-SepClassification model=DPN1312020.02 | 55.72 | 63.64 | — | |
| SLTLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 55.7 | 65.4 | 67.6 | |
| PSOL-DenseNet161-SepBackbone=DenseNet161, Parameters=56.29M, FLOPs=15.46G2020.02 | 55.31 | 64.18 | 66.28 | |
| FAMLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 55.2 | — | 68.6 | |
| SPGClassification model=DPN1312020.02 | 55.19 | 62.76 | — | |
| VGG16-ACOLClassification model=DPN-ensemble2020.02 | 54.86 | 61.45 | — | |
| I2CBackbone=ResNet502022.04 | 54.83 | 64.6 | 68.5 | |
| PSOL-InceptionV3-SepBackbone=InceptionV3, Parameters=53.32M, FLOPs=11.42G2020.02 | 54.82 | 63.25 | 65.21 | |
| FAMLoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 54.5 | — | 64.6 | |
| PSOL-DenseNet161-JointBackbone=DenseNet161, Parameters=29.81M, FLOPs=7.80G2020.02 | 54.48 | 63.41 | 65.39 | |
| PSOL-ResNet50-SepBackbone=ResNet50, Parameters=50.12M, FLOPs=8.18G2020.02 | 53.98 | 63.08 | 65.44 | |
| VGG16-ACOLClassification model=DPN1312020.02 | 53.94 | 61.15 | — | |
| OursBackbone=ResNet502022.04 | 53.76 | 65.75 | 69.89 | |
| DGLBackbone=ResNet502022.04 | 53.41 | 62.69 | 69.34 | |
| TAFormerLoc Back.=VGG16, Cls Back.=VGG162023.07 | 53.4 | 67.7 | 74 | |
| TS-CAMLoc Back.=Deit-S, Cls Back.=Deit-S2023.07 | 53.4 | 64.3 | 67.6 | |
| BASLoc Back.=VGG16, Cls Back.=VGG162023.07 | 53 | 65.4 | 69.6 | |
| BASLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 53 | 66.6 | 72 | |
| PSOL-ResNet50-JointBackbone=ResNet50, Parameters=26.61M, FLOPs=4.10G2020.02 | 52.82 | 62 | 64.3 | |
| PSOL-InceptionV3-JointBackbone=InceptionV3, Parameters=29.21M, FLOPs=5.71G2020.02 | 52.76 | 61.1 | 62.83 | |
| FAMLoc Back.=VGG16, Cls Back.=VGG162023.07 | 52 | — | 71.7 | |
| SLTLoc Back.=VGG16, Cls Back.=VGG162023.07 | 51.2 | 62.4 | 67.2 | |
| PSOL-VGG16-SepBackbone=VGG16, Parameters=274.72M, FLOPs=30.83G2020.02 | 50.89 | 60.9 | 64.03 | |
| OursBackbone=VGG162022.04 | 49.94 | 63.25 | 68.92 | |
| Pan et al.Backbone=VGG162022.04 | 49.56 | 61.32 | 65.05 | |
| Bae et al.Backbone=ResNet502022.04 | 49.42 | — | 62.2 | |
| PSOL-VGG16-JointBackbone=VGG16, Parameters=140.46M, FLOPs=15.42G2020.02 | 48.83 | 59 | 62.1 | |
| ADLBackbone=InceptionV3, Parameters=38.45M, FLOPs=66.59G2020.02 | 48.71 | — | — | |
| SPGBackbone=InceptionV3, Parameters=38.45M, FLOPs=66.59G2020.02 | 48.6 | 60 | 64.69 | |
| ADLBackbone=ResNet50-SE, Parameters=28.09M, FLOPs=6.10G2020.02 | 48.53 | — | — | |
| Ki et al.Backbone=ResNet502022.04 | 48.4 | — | — | |
| PSOL-VGG-GAP-SepBackbone=VGG-GAP, Parameters=29.64M, FLOPs=30.70G2020.02 | 48.36 | 58.75 | 63.72 | |
| ADLBackbone=ResNet502022.04 | 48.23 | — | 61.04 | |
| DGLBackbone=VGG162022.04 | 47.66 | 58.89 | 64.78 | |
| TAFormerLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 47.6 | 65.5 | 68.8 | |
| DANetLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 47.5 | 58.3 | — | |
| I2CBackbone=VGG162022.04 | 47.41 | 58.51 | 63.9 | |
| PSOL-VGG-GAP-JointBackbone=VGG-GAP, Parameters=15.08M, FLOPs=15.35G2020.02 | 47.37 | 58.41 | 62.25 | |
| DDT-VGG16Backbone=VGG16, Parameters=138.36M, FLOPs=15.42G2020.02 | 47.31 | 58.23 | 61.41 | |
| CutMixBackbone=ResNet50, Parameters=26.61M, FLOPs=4.10G2020.02 | 47.25 | — | — | |
| CutMixBackbone=ResNet502022.04 | 47.25 | — | — | |
| Ki et al.Backbone=VGG162022.04 | 47.2 | — | — | |
| MEILBackbone=VGG162022.04 | 46.81 | — | — | |
| CAMLoc Back.=InceptionV3, Cls Back.=InceptionV32023.07 | 46.3 | 58.2 | 62.7 | |
| EILBackbone=VGG162022.04 | 46.27 | — | — | |
| FAMLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 46.2 | — | 62.1 | |
| VGG16-ACoLBackbone=VGG-GAP, Parameters=45.08M, FLOPs=43.32G2020.02 | 45.83 | 59.43 | 62.96 | |
| ACOLBackbone=VGG162022.04 | 45.83 | 59.43 | 62.96 | |
| ADLBackbone=VGG-GAP, Parameters=14.82M, FLOPs=15.35G2020.02 | 44.92 | — | — | |
| ADLBackbone=VGG162022.04 | 44.92 | — | — | |
| Bae et al.Backbone=VGG162022.04 | 44.62 | — | 60.73 | |
| VGG16-Grad-CAMBackbone=VGG16, Parameters=138.36M, FLOPs=15.42G2020.02 | 43.49 | 53.59 | — | |
| CutMixBackbone=VGG-GAP, Parameters=138.36M, FLOPs=15.35G2020.02 | 43.45 | — | — | |
| CutMixBackbone=VGG162022.04 | 43.45 | — | — | |
| CAMLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 43.4 | 54.4 | 59 | |
| ADLLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 43 | — | — | |
| VGG16-CAMBackbone=VGG-GAP, Parameters=14.82M, FLOPs=15.35G2020.02 | 42.8 | 54.86 | 59 | |
| CAMBackbone=VGG162022.04 | 42.8 | 54.86 | — | |
| CAMLoc Back.=VGG16, Cls Back.=VGG162023.07 | 42.8 | 54.9 | 59 | |
| HaSLoc Back.=MobileNetV1, Cls Back.=MobileNetV12023.07 | 42.7 | — | 60.1 | |
| DenseNet161-CAMBackbone=DenseNet161, Parameters=29.81M, FLOPs=7.80G2020.02 | 39.61 | 50.4 | 52.54 | |
| CAMLoc Back.=ResNet50, Cls Back.=ResNet502023.07 | 39 | 49.5 | 51.9 | |
| ResNet50-CAMBackbone=ResNet50, Parameters=25.56M, FLOPs=4.10G2020.02 | 38.99 | 49.47 | 51.86 | |
| ADLLoc Back.=ResNet50-SE, Cls Back.=ResNet50-SE2023.07 | — | — | 48.5 | |
| AlexNetsupervision=full2020.02 | — | 65.8 | — | |
| DenseNet161-Sepsupervision=weak2020.02 | — | 64.2 | — | |
| GoogLeNetsupervision=full2020.02 | — | 73.3 | — | |
| GoogLeNet-GAPsupervision=weak2020.02 | — | 57.1 | — | |
| GoogLeNet-GAP (heuristics)supervision=weak2020.02 | — | 62.9 | — | |
| OverFeatsupervision=full2020.02 | — | 70.1 | — | |
| ResNet + Faster-RCNN-ensemblesupervision=full2020.02 | — | 90 | — | |
| VGG16supervision=full2020.02 | — | 70.5 | — | |
| VGG16-Sepsupervision=weak2020.02 | — | 60.9 | — | |
| VGGNet-ensemblesupervision=full2020.02 | — | 73.1 | — |