TDFNet: An Efficient Audio-Visual Speech Separation Model with Top-down Fusion

Benchmark Model Rank Results
speech-separation-on-lrs2TDFNet-large#2SI-SNRi: 15.8SDRi: 15.9PESQ: 3.21STOI: 0.949
speech-separation-on-lrs2TDFNet (MHSA + Shared)#3SI-SNRi: 15.0SDRi: 15.2PESQ: 3.16STOI: 0.938
speech-separation-on-lrs2TDFNet-small#8SI-SNRi: 13.6SDRi: 13.7PESQ: 3.10STOI: 0.931