ModelECAPA-TDNN
ECAPA
A TDNN-based speaker embedding model widely used for speaker verification and speaker diarization tasks.
Timeline (last 90 days)
Nov 16
Marimba's speaker-change detection achieves AUROC of 0.966–0.980 on AMI and VoxConverse with the layer-12 readout, outperforming the ECAPA baseline.
Unverified50%