Valence-Arousal Estimation on Aff-wild (test)
0.628CCC (V)ELIM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ELIMBackbone=AL-tuned, Number of parameters=2.6M2022.09 | 0.628 | 0.228 | 0.199 | 0.634 | 0.638 | 0.611 | |
| ELIMBackbone=Mlp-Mixer, Number of parameters=59.1M2022.09 | 0.611 | 0.214 | 0.197 | 0.645 | 0.684 | 0.631 | |
| ELIMBackbone=ResNet18, Number of parameters=11.7M2022.09 | 0.609 | 0.217 | 0.197 | 0.635 | 0.602 | 0.57 | |
| MIMAMOBackbone=ResNet50+GRU, Number of parameters=>25.6M2022.09 | 0.58 | — | — | — | — | 0.52 | |
| CAFBackbone=ResNet18, Number of parameters=11M2022.09 | 0.55 | 0.22 | 0.2 | 0.57 | 0.57 | 0.56 | |
| CAFBackbone=AL-tuned, Number of parameters=2.7M2022.09 | 0.54 | 0.24 | 0.21 | 0.55 | 0.57 | 0.56 | |
| BreG-NeXtBackbone=ResNeXt50, Number of parameters=3.1M2022.09 | 0.37 | 0.26 | 0.31 | 0.42 | 0.4 | 0.31 | |
| Hasani et al.Backbone=Inception-ResNet, Number of parameters=~56M2022.09 | 0.36 | 0.27 | 0.36 | 0.44 | 0.26 | 0.19 |