search

TaoRuijie / ECAPA-TDNN

Unofficial reimplementation of ECAPA-TDNN for speaker recognition (EER=0.86 for Vox1_O when train only in Vox2)
MIT License
594 stars 113 forks source link

算法正确率并不高？ #39

Closed JJun-Guo closed 2 years ago

JJun-Guo commented 2 years ago

80 epoch, LR 0.000090, LOSS 1.371865, ACC 74.04%, EER 1.03%, bestEER 0.96% 看到训练log，在第80个epoch的时候，eer已经达到1.03，但是ACC却只有74.04，正确率并不高呢？

TaoRuijie commented 2 years ago

请查看aamsoftmax算法和softmax算法的不同