Action Unit Detection on BP4D (5-fold cross-val)
68.5Average PerformanceSSM
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SSMBackbone=CLIP-ViT-B/16, Learning Protocol=Joint Learning2026.04 | 68.5 | 61 | 48.4 | 56 | 81.8 | 83.1 | 84.9 | 88.7 | 70.7 | 59.5 | 68 | 60.1 | 59.8 | |
| FLCMBackbone=ResNet-50, Learning Protocol=Joint Learning2026.04 | 67.7 | 60.6 | 50.3 | 64.2 | 80.7 | 80.5 | 85.9 | 88.6 | 68 | 57.3 | 63.4 | 52 | 60.5 | |
| MAE-FaceBackbone=ViT-B/16, Learning Protocol=Joint Learning2026.04 | 67.4 | 62.5 | 56.4 | 66.3 | 79.6 | 79.6 | 85.6 | 89.1 | 64.2 | 54.5 | 65 | 53.8 | 51.8 | |
| CausalAffectBackbone=ResNet-50, Learning Protocol=Joint Learning2026.04 | 67.4 | 67.1 | 43.6 | 66 | 80.1 | 79.1 | 84.8 | 88.9 | 71.1 | 55.6 | 66.6 | 47.5 | 58.8 | |
| FMAE-IATBackbone=ViT-L/16, Learning Protocol=Joint Learning2026.04 | 67.1 | 62.7 | 51.9 | 62.7 | 79.8 | 80.1 | 84.8 | 89.9 | 64.6 | 54.9 | 65.4 | 53.1 | 54.7 | |
| MDHRBackbone=Swin-B, Learning Protocol=Joint Learning2026.04 | 66.6 | 58.3 | 50.9 | 58.9 | 78.4 | 80.3 | 84.9 | 88.2 | 69.5 | 56 | 65.5 | 49.5 | 59.3 | |
| FMAEBackbone=ViT-L/16, Learning Protocol=Joint Learning2026.04 | 66.6 | 59.2 | 50 | 62.7 | 80 | 79.2 | 84.7 | 89.8 | 63.5 | 52.8 | 65.1 | 55.3 | 56.9 | |
| HiVABackbone=Swin-B, Learning Protocol=Joint Learning2026.04 | 66.5 | 54.3 | 49.7 | 63.3 | 79.3 | 79.8 | 84.5 | 88.8 | 68.5 | 57 | 62.6 | 53.1 | 56.8 | |
| AUFormerBackbone=ViT-B/16, Learning Protocol=Joint Learning2026.04 | 66.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| STLBackbone=CLIP-ViT-B/16, Learning Protocol=Single-Task2026.04 | 66.2 | 58.8 | 48 | 60.5 | 78.4 | 79.6 | 84.3 | 88.8 | 69.5 | 51.7 | 65.7 | 53.1 | 55.7 | |
| CLEFBackbone=CLIP-ViT-B/16, Learning Protocol=Joint Learning2026.04 | 65.9 | 55.8 | 46.8 | 63.3 | 79.5 | 77.6 | 83.6 | 87.8 | 67.3 | 55.2 | 63.5 | 53 | 57.8 | |
| AU-TTTBackbone=ViT-S/16, Learning Protocol=Joint Learning2026.04 | 65.6 | — | — | — | — | — | — | — | — | — | — | — | — | |
| KSBackbone=ResNet-18, Learning Protocol=Joint Learning2026.04 | 64.7 | 55.3 | 48.6 | 57.1 | 77.5 | 81.8 | 83.3 | 86.4 | 62.8 | 52.3 | 61.3 | 51.6 | 58.3 | |
| KSRLBackbone=ResNet-50, Learning Protocol=Joint Learning2026.04 | 64.5 | 53.3 | 47.4 | 56.2 | 79.4 | 80.7 | 85.1 | 89 | 67.4 | 55.9 | 61.9 | 48.5 | 49 |