speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity
ebdf631d98
FunASR/egs/cnceleb/resnet34
History
仁迷 8d8c77d0b6 modify aishell transformer directory name
2023-02-01 17:47:44 +08:00
..
local update version 0.1.6 2023-01-16 18:46:40 +08:00
sid update version 0.1.6 2023-01-16 18:46:40 +08:00
configuration.json update version 0.1.6 2023-01-16 18:46:40 +08:00
path.sh update version 0.1.6 2023-01-16 18:46:40 +08:00
README.md update version 0.1.6 2023-01-16 18:46:40 +08:00
run.sh update version 0.1.6 2023-01-16 18:46:40 +08:00
sv.yaml update version 0.1.6 2023-01-16 18:46:40 +08:00
utils modify aishell transformer directory name 2023-02-01 17:47:44 +08:00

README.md

ResNet34 Result

Training Config

  • Feature info: using 80 dims fbank, no cmvn, speed perturb(0.9, 1.0, 1.1)
  • Train info: lr 1e-4, batch_size 64, 1 gpu(Tesla V100), acc_grad 1, 300000 steps, clip_gradient_norm 3.0, weight_l2_regularizer 0.01
  • Loss info: additive angular margin softmax, feature_scaling_factor=8, margin 0.25
  • Model info: ResNet34, global statistics pooling, Dense
  • Train config: conf/train_sv_resnet34.yaml
  • Model size: 5.60 M parameters

Results (EER & minDCF)

  • Test set: Alimeeting-test, CN-Celeb-eval-speech
testset EER(%) minDCF Threshold
Alimeeting-test 1.45 0.0849 0.9666
CN-Celeb-eval-speech 9.00 0.2936 0.9465