wq2012/univ_conformer_tiny_ge2e
Audio Classification • 693k • Updated
FlaxSpeaker
Open-Source Speaker Recognition & Verification Models in JAX/Flax (Code: https://github.com/wq2012/FlaxSpeaker)
Note conformer (tiny) | atp | ge2e_softmax | cosine | train=universal — LS 13.23%, CN 22.00%, VOX 22.55%, Pooled 1-utt 18.75% / 3-utt 15.03% EER
wq2012/univ_conformer_small_ge2e
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_softmax | cosine | train=universal + FP32/INT8 TFLite — LS 12.50%, CN 22.33%, VOX 24.80%, Pooled 1-utt 19.40% / 3-utt 16.03% EER
wq2012/univ_conformer_small_dr_vectors
Audio Classification • 2.09M • UpdatedNote conformer (small) | atp | extended_set_softmax | dr_vector | train=universal — LS 15.20%, CN 22.33%, VOX 22.75%, Pooled 1-utt 20.10% / 3-utt 15.40% EER
wq2012/univ_conformer_small_pfas_ge2e
Audio Classification • 2.16M • UpdatedNote conformer (small) | pfas | ge2e_softmax | pfas | train=universal — LS 15.20%, CN 21.93%, VOX 23.65%, Pooled 1-utt 19.86% / 3-utt 16.27% EER
wq2012/univ_conformer_small_pfas
Audio Classification • 2.16M • UpdatedNote conformer (small) | pfas | extended_set_softmax | pfas | train=universal — LS 15.83%, CN 23.17%, VOX 22.95%, Pooled 1-utt 20.35% / 3-utt 16.00% EER
wq2012/univ_conformer_small_ext_softmax
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | extended_set_softmax | cosine | train=universal — LS 14.83%, CN 22.40%, VOX 23.30%, Pooled 1-utt 20.50% / 3-utt 16.17% EER
wq2012/univ_mamba_small_ge2e
Audio Classification • 1.03M • UpdatedNote mamba (small) | asp | ge2e_softmax | cosine | train=universal — LS 14.20%, CN 23.20%, VOX 23.75%, Pooled 1-utt 20.41% / 3-utt 16.07% EER
wq2012/univ_transformer_small_ge2e
Audio Classification • 1.24M • UpdatedNote transformer (small) | atp | ge2e_softmax | cosine | train=universal — LS 15.17%, CN 22.67%, VOX 23.95%, Pooled 1-utt 20.00% / 3-utt 16.33% EER
wq2012/univ_lstm_small_ge2e
Audio Classification • 1.71M • UpdatedNote lstm (small) | asp | ge2e_softmax | cosine | train=universal + FP32/INT8 TFLite — LS 14.97%, CN 25.53%, VOX 22.50%, Pooled 1-utt 21.31% / 3-utt 18.00% EER
wq2012/univ_ecapa_tdnn_tiny_ge2e
Audio Classification • 627k • UpdatedNote ecapa_tdnn (tiny) | asp | ge2e_softmax | cosine | train=universal — LS 19.07%, CN 25.27%, VOX 23.35%, Pooled 1-utt 23.02% / 3-utt 18.00% EER
wq2012/univ_ecapa_tdnn_small_ge2e
Audio Classification • 1.68M • UpdatedNote ecapa_tdnn (small) | asp | ge2e_softmax | cosine | train=universal + FP32/INT8 TFLite — LS 19.00%, CN 25.13%, VOX 23.30%, Pooled 1-utt 23.16% / 3-utt 18.00% EER
wq2012/univ_ecapa_tdnn_base_ge2e
Audio Classification • 3.19M • UpdatedNote ecapa_tdnn (base) | asp | ge2e_softmax | cosine | train=universal — LS 18.63%, CN 25.97%, VOX 22.20%, Pooled 1-utt 22.81% / 3-utt 18.20% EER
wq2012/univ_conformer_base_ge2e
Audio Classification • 4.55M • UpdatedNote conformer (base) | atp | ge2e_softmax | cosine | train=universal — LS 15.37%, CN 22.40%, VOX 22.85%, Pooled 1-utt 19.85% / 3-utt 16.80% EER
wq2012/univ_resnet_small_ge2e
Audio Classification • 2.24M • UpdatedNote resnet (small) | asp | ge2e_softmax | cosine | train=universal — LS 24.53%, CN 29.33%, VOX 28.25%, Pooled 1-utt 27.89% / 3-utt 25.47% EER
wq2012/univ_conformer_small_triplet
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | triplet | cosine | train=universal — LS 18.50%, CN 26.67%, VOX 25.40%, Pooled 1-utt 24.07% / 3-utt 20.33% EER
wq2012/univ_conformer_small_arcface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | arcface | cosine | train=universal — LS 20.57%, CN 25.27%, VOX 25.00%, Pooled 1-utt 25.11% / 3-utt 21.67% EER
wq2012/univ_conformer_small_cosface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | cosface | cosine | train=universal — LS 20.37%, CN 25.73%, VOX 26.40%, Pooled 1-utt 25.60% / 3-utt 21.83% EER
wq2012/univ_conformer_small_sphereface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | sphereface | cosine | train=universal — LS 22.27%, CN 25.67%, VOX 25.15%, Pooled 1-utt 25.84% / 3-utt 22.07% EER
wq2012/univ_conformer_small_ge2e_contrast
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_contrast | cosine | train=universal — LS 20.67%, CN 31.77%, VOX 26.65%, Pooled 1-utt 27.24% / 3-utt 20.37% EER
wq2012/cn_conformer_small_dr_vectors
Audio Classification • 2.09M • UpdatedNote conformer (small) | atp | extended_set_softmax | dr_vector | train=cnceleb2 — LS 23.00%, CN 21.17%, VOX 26.30%, Pooled 1-utt 23.95% / 3-utt 18.97% EER
wq2012/cn_conformer_small_ge2e
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_softmax | cosine | train=cnceleb2 — LS 20.80%, CN 21.33%, VOX 27.20%, Pooled 1-utt 23.93% / 3-utt 20.13% EER
wq2012/cn_conformer_small_pfas
Audio Classification • 2.16M • UpdatedNote conformer (small) | pfas | extended_set_softmax | pfas | train=cnceleb2 — LS 24.47%, CN 22.13%, VOX 28.15%, Pooled 1-utt 26.20% / 3-utt 23.20% EER
wq2012/cn_ecapa_tdnn_small_arcface
Audio Classification • 1.68M • UpdatedNote ecapa_tdnn (small) | asp | arcface | cosine | train=cnceleb2 — LS 30.53%, CN 28.07%, VOX 31.40%, Pooled 1-utt 30.23% / 3-utt 26.47% EER
wq2012/ls_conformer_base_ge2e
Audio Classification • 4.55M • UpdatedNote conformer (base) | atp | ge2e_softmax | cosine | train=librispeech — LS 10.77%, CN 28.87%, VOX 24.65%, Pooled 1-utt 21.26% / 3-utt 17.00% EER
wq2012/ls_conformer_small_arcface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | arcface | cosine | train=librispeech — LS 14.47%, CN 28.73%, VOX 23.85%, Pooled 1-utt 22.80% / 3-utt 18.23% EER
wq2012/ls_conformer_small_cosface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | cosface | cosine | train=librispeech — LS 14.13%, CN 30.53%, VOX 23.00%, Pooled 1-utt 23.26% / 3-utt 18.17% EER
wq2012/ls_conformer_small_dr_vectors
Audio Classification • 2.09M • UpdatedNote conformer (small) | atp | extended_set_softmax | dr_vector | train=librispeech — LS 11.17%, CN 28.80%, VOX 22.30%, Pooled 1-utt 20.69% / 3-utt 16.60% EER
wq2012/ls_conformer_small_ext_softmax
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | extended_set_softmax | cosine | train=librispeech — LS 10.03%, CN 29.87%, VOX 24.10%, Pooled 1-utt 21.05% / 3-utt 17.53% EER
wq2012/ls_conformer_small_ge2e
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_softmax | cosine | train=librispeech + FP32/INT8 TFLite — LS 9.67%, CN 28.50%, VOX 22.00%, Pooled 1-utt 20.12% / 3-utt 16.73% EER
wq2012/ls_conformer_small_ge2e_contrast
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_contrast | cosine | train=librispeech — LS 22.60%, CN 36.30%, VOX 32.50%, Pooled 1-utt 31.93% / 3-utt 26.43% EER
wq2012/ls_conformer_small_pfas
Audio Classification • 2.16M • UpdatedNote conformer (small) | pfas | extended_set_softmax | pfas | train=librispeech — LS 10.50%, CN 28.83%, VOX 22.90%, Pooled 1-utt 20.88% / 3-utt 16.80% EER
wq2012/ls_conformer_small_pfas_ge2e
Audio Classification • 2.16M • UpdatedNote conformer (small) | pfas | ge2e_softmax | pfas | train=librispeech — LS 10.50%, CN 28.87%, VOX 23.40%, Pooled 1-utt 20.69% / 3-utt 17.17% EER
wq2012/ls_conformer_small_sphereface
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | sphereface | cosine | train=librispeech — LS 14.73%, CN 29.93%, VOX 22.20%, Pooled 1-utt 23.29% / 3-utt 17.97% EER
wq2012/ls_conformer_small_triplet
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | triplet | cosine | train=librispeech — LS 16.07%, CN 28.93%, VOX 22.00%, Pooled 1-utt 23.41% / 3-utt 19.80% EER
wq2012/ls_conformer_tiny_ge2e
Audio Classification • 693k • UpdatedNote conformer (tiny) | atp | ge2e_softmax | cosine | train=librispeech — LS 10.53%, CN 30.00%, VOX 25.70%, Pooled 1-utt 21.91% / 3-utt 18.07% EER
wq2012/ls_ecapa_tdnn_base_ge2e
Audio Classification • 3.19M • UpdatedNote ecapa_tdnn (base) | asp | ge2e_softmax | cosine | train=librispeech — LS 14.47%, CN 33.33%, VOX 19.90%, Pooled 1-utt 23.29% / 3-utt 18.90% EER
wq2012/ls_ecapa_tdnn_small_ge2e
Audio Classification • 1.68M • UpdatedNote ecapa_tdnn (small) | asp | ge2e_softmax | cosine | train=librispeech + FP32/INT8 TFLite — LS 13.83%, CN 31.70%, VOX 19.65%, Pooled 1-utt 22.48% / 3-utt 17.83% EER
wq2012/ls_ecapa_tdnn_tiny_ge2e
Audio Classification • 627k • UpdatedNote ecapa_tdnn (tiny) | asp | ge2e_softmax | cosine | train=librispeech — LS 12.60%, CN 30.87%, VOX 21.70%, Pooled 1-utt 22.10% / 3-utt 17.67% EER
wq2012/ls_lstm_small_ge2e
Audio Classification • 1.71M • UpdatedNote lstm (small) | asp | ge2e_softmax | cosine | train=librispeech + FP32/INT8 TFLite — LS 12.77%, CN 28.03%, VOX 23.35%, Pooled 1-utt 21.25% / 3-utt 17.73% EER
wq2012/ls_mamba_small_ge2e
Audio Classification • 1.03M • UpdatedNote mamba (small) | asp | ge2e_softmax | cosine | train=librispeech — LS 12.17%, CN 29.13%, VOX 21.05%, Pooled 1-utt 21.30% / 3-utt 17.10% EER
wq2012/ls_resnet_small_ge2e
Audio Classification • 2.24M • UpdatedNote resnet (small) | asp | ge2e_softmax | cosine | train=librispeech — LS 15.20%, CN 32.17%, VOX 21.40%, Pooled 1-utt 23.94% / 3-utt 20.47% EER
wq2012/ls_transformer_small_ge2e
Audio Classification • 1.24M • UpdatedNote transformer (small) | atp | ge2e_softmax | cosine | train=librispeech — LS 10.60%, CN 28.10%, VOX 22.30%, Pooled 1-utt 20.80% / 3-utt 16.37% EER
wq2012/vox_conformer_small_ge2e
Audio Classification • 2.03M • UpdatedNote conformer (small) | atp | ge2e_softmax | cosine | train=voxceleb1 — LS 16.33%, CN 27.87%, VOX 20.80%, Pooled 1-utt 22.40% / 3-utt 17.63% EER
wq2012/vox_ecapa_tdnn_small_ext_softmax
Audio Classification • 1.68M • UpdatedNote ecapa_tdnn (small) | asp | extended_set_softmax | cosine | train=voxceleb1 — LS 20.07%, CN 32.40%, VOX 21.10%, Pooled 1-utt 24.57% / 3-utt 20.97% EER