Weakly Supervised Object Localization on ImageNet-1K (test)
68.7MaxBoxAccV2 (Mean)Feature Direction Alignment + Consistency with Attentive Dropout
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Feature Direction Alignment + Consistency with Attentive DropoutBackbone=ResNet502022.04 | 68.7 | 86.7 | 71.1 | 48.3 | |
| Feature Direction Alignment + Consistency with Attentive DropoutBackbone=VGG162022.04 | 66.6 | 84.8 | 69.2 | 45.9 | |
| Ki et al.Backbone=ResNet502022.04 | 65.2 | 84.3 | 67.6 | 43.6 | |
| ADL + IVRBackbone=ResNet502022.04 | 65.1 | — | — | — | |
| HaS + PaSBackbone=ResNet502022.04 | 64.6 | — | — | — | |
| CAMBackbone=ResNet502022.04 | 63.7 | 83.7 | 65.7 | 41.6 | |
| ADLBackbone=ResNet502022.04 | 63.7 | 83.6 | 65.6 | 41.8 | |
| ADL + IVRBackbone=VGG162022.04 | 63.7 | — | — | — | |
| HaSBackbone=ResNet502022.04 | 63.4 | 83.7 | 65.2 | 41.3 | |
| CALMBackbone=ResNet502022.04 | 63.4 | — | — | — | |
| SPGBackbone=ResNet502022.04 | 63.3 | 83.9 | 65.4 | 40.6 | |
| CutMixBackbone=ResNet502022.04 | 63.3 | 83.7 | 65.2 | 41 | |
| CALMBackbone=VGG162022.04 | 62.8 | — | — | — | |
| HaS + PaSBackbone=VGG162022.04 | 62.1 | — | — | — | |
| Ki et al.Backbone=VGG162022.04 | 61.3 | 81.5 | 63.2 | 39.4 | |
| HaSBackbone=VGG162022.04 | 60.6 | 80.7 | 62.1 | 38.9 | |
| CAMBackbone=VGG162022.04 | 60 | 81 | 62 | 37.1 | |
| SPGBackbone=VGG162022.04 | 59.9 | 81.4 | 62 | 36.3 | |
| ADLBackbone=VGG162022.04 | 59.9 | 80.8 | 60.9 | 37.8 | |
| CutMixBackbone=VGG162022.04 | 59.5 | 80.3 | 61 | 37.1 |