Environmental Sound Deepfake Detection on EnvSDD Track 2 1.0 (Final ranking-phase)
35EEREnsemble
Evaluation results
| Method | Links | |
|---|---|---|
| EnsembleEnsemble components=0.2×BEATs-AASIST + 0.5×BEAT2AASIST (Freq., CNN-gate, k=4) + 0.3×BEAT2AASIST (Freq., CNN-gate, k=10)2025.12 | 35 | |
| EnsembleEnsemble components=0.4×BEATs-AASIST + 0.6×BEAT2AASIST (Freq., CNN-gate, k=4)2025.12 | 40 | |
| BEAT2AASISTFusion strategy=Frequency-based, Gate type=CNN-gate, k (number of layers)=42025.12 | 46 | |
| BEAT2AASISTFusion strategy=Frequency-based, Gate type=CNN-gate, k (number of layers)=102025.12 | 60 | |
| BEATs-AASISTFusion strategy=Concatenation, k (number of layers)=42025.12 | 75 |