Active Speaker Detection on Columbia
0.981F1 (Bell)UniCon
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| UniConModality=Visual-only, Zero-shot=true2021.08 | 0.981 | 0.879 | 0.979 | 0.993 | 0.992 | 0.965 | |
| SyncNetModality=Audio-visual, Zero-shot=true, Large-scale self-supervised pre-training=true2021.08 | 0.937 | 0.834 | 0.868 | 0.977 | 0.861 | 0.895 | |
| UniConModality=Audio-visual, Zero-shot=true2021.08 | 0.936 | 0.813 | 0.938 | 0.935 | 0.921 | 0.909 | |
| LWTNetModality=Audio-visual, Zero-shot=true, Large-scale self-supervised pre-training=true2021.08 | 0.926 | 0.824 | 0.887 | 0.944 | 0.959 | 0.908 | |
| S-VVADModality=Visual-only, Zero-shot=true2021.08 | 0.924 | 0.972 | 0.923 | 0.955 | 0.925 | 0.94 | |
| Upper BodyModality=Visual-only, Zero-shot=true2021.08 | 0.873 | 0.964 | 0.922 | 0.83 | 0.872 | 0.892 | |
| Cross-modalModality=Visual-only, Zero-shot=true2021.08 | 0.829 | 0.658 | 0.736 | 0.869 | 0.818 | 0.802 |