Active Speaker Detection on RealVAD
96.47P1UniCon
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| UniConModality=Visual, Evaluation Protocol=fine-tuned2021.08 | 96.47 | 81.05 | 86.93 | 84.36 | 89.94 | 85.6 | 94.92 | 88.07 | 90.9 | 88.69 | 4.66 | |
| UniConModality=Visual, Evaluation Protocol=zero-shot2021.08 | 94.32 | 73.98 | 89.93 | 76.73 | 80.57 | 93.61 | 98.78 | 83.51 | 93.54 | 87.22 | 8.27 | |
| UniConModality=Audio-Visual, Evaluation Protocol=fine-tuned2021.08 | 86.87 | 76.45 | 81.55 | 86.98 | 79.58 | 88.86 | 97.01 | 84.5 | 88.92 | 85.64 | 5.7 | |
| UniConModality=Audio-Visual, Evaluation Protocol=zero-shot2021.08 | 86.72 | 78.11 | 70.48 | 73.14 | 68.93 | 84.9 | 93.01 | 80.4 | 86.99 | 80.3 | 7.82 | |
| BeyanModality=Visual, Evaluation Protocol=not specified2021.08 | 51.63 | 53.49 | 42.92 | 51.7 | 44.4 | 50.48 | 58.73 | 67.94 | 55.75 | 53.04 | 7.5 |