VoxCeleb
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
VoxCeleb (test)
98.23FID
4
VoxCeleb
75.642FLOPs (G)
4
Voxceleb-3D
1.52Error Rate (ER)
4
VoxCeleb (test)
0.04L1 Loss
4
VoxCeleb-Sim
9.31SI-SDR (dB)
3
VoxCeleb2 with FSD50K environmental noise Scenario 1 (test)
11.41SI-SNRi
3
VoxCeleb2 Scenario 3: 1 speaker + environmental noise + 2 interfering speakers (test)
5.37SI-SNRi
3
VoxCeleb2 Scenario 4: 1 speaker + music noise + 2 interfering speakers
4.11SI-SNRi
3
VoxCeleb2 Scenario 2: 1 speaker + music noise from FMA
7.49SI-SNRi
3
VoxCeleb2 (test)
3.53UTMOS
3
VoxCeleb
0.869Similarity to Source
3
VoxCeleb
0.838Source Similarity
3
VoxCeleb
0.926Sim to Source
3
VoxCeleb2 (held-out)
3.84Predicted K
2
VoxCeleb2 (test)
3.448UTMOS
2
VoxCeleb 2
0.022Identity Preservation
2
Voxceleb2 (Sequences)
29.15PSNR
2
Voxceleb2 Single Images
29.22PSNR
2
VoxCeleb subset
9.341FID (Age Edit)
2
VoxCeleb
38.126PSNR
2
VoxCeleb1
66User Preference Score
2
VoxCeleb2
8.86FID
2
VoxCeleb2 (test)
16.51FID
2
Multimodal VoxCeleb
0.197CLIP Score
2
VoxCeleb (test)
0.299Accuracy
2