Multi-label HOI classification on HICO
36.1mAPFusion, weighted loss
Evaluation Results
| Method | Links | |
|---|---|---|
| Fusion, weighted lossFull Im.=true, Bbox/Pose=true, MIL=true, Wtd Loss=true2017.11 | 36.1 | |
| Attentional PoolingFull Im.=true, Bbox/Pose=false, MIL=false, Wtd Loss=false, Backbone=ResNet-1012017.11 | 35 | |
| Pose Regularized Attentional PoolingFull Im.=true, Bbox/Pose=true, MIL=false, Wtd Loss=false, Backbone=ResNet-1012017.11 | 34.6 | |
| FusionFull Im.=true, Bbox/Pose=true, MIL=true, Wtd Loss=false2017.11 | 33.8 | |
| ResNet101, full imageFull Im.=true, Bbox/Pose=false, MIL=false, Wtd Loss=false, Backbone=ResNet-1012017.11 | 30.2 | |
| VGG16, full imageFull Im.=true, Bbox/Pose=false, MIL=false, Wtd Loss=false2017.11 | 29.4 | |
| Scene-RCNNFull Im.=true, Bbox/Pose=true, MIL=true, Wtd Loss=false2017.11 | 29 | |
| R*CNNFull Im.=true, Bbox/Pose=true, MIL=true, Wtd Loss=false2017.11 | 28.5 | |
| ResNet101 with CBPFull Im.=true, Bbox/Pose=false, MIL=false, Wtd Loss=false, Backbone=ResNet-1012017.11 | 26.8 | |
| AlexNet+SVMFull Im.=true, Bbox/Pose=false, MIL=false, Wtd Loss=false2017.11 | 19.4 |