OpenVoiceOS/redimnet-b2-vox2-onnx

Updated 20 days ago

Note ReDimNet b2 — compact high-accuracy speaker embedding (1.8M params). Trained on VoxCeleb2, exported to ONNX. 192-dim embedding, raw audio input.

Wespeaker/wespeaker-voxceleb-resnet34-LM

Summarization • Updated May 6, 2024 • 110 • 9

Note WeSpeaker ResNet34, 256-dim, fbank80 — default model.

Wespeaker/wespeaker-ecapa-tdnn512-LM

Summarization • Updated May 6, 2024 • 34 • 2

Note WeSpeaker ECAPA-TDNN-512, 192-dim, fbank80.

Wespeaker/wespeaker-voxceleb-resnet293-LM

Summarization • Updated May 6, 2024 • 133 • 5

Note WeSpeaker ResNet293, 256-dim, 28.6M params — highest accuracy.

csukuangfj/speaker-embedding-models

Updated Oct 14, 2024 • 5

Note Bundle: WeSpeaker CAM++, 3D-Speaker CAM++, ERes2Net, TitaNet, Speakernet.