search

astorfi / 3D-convolutional-speaker-recognition

:speaker: Deep Learning & 3D Convolutional Neural Networks for Speaker Verification
Apache License 2.0
778 stars 275 forks source link

Testing accuracy is not increasing #31

Closed AbhishekVagrecha closed 6 years ago

AbhishekVagrecha commented 6 years ago

Hi,

I am getting a training accuracy of 95% on voxceleb but testing accuracy is around 10% only. What can be reason of this? Speakers=1211 Batch size =100 epoch = 50 Data size = 24000

AbhishekVagrecha commented 6 years ago

And how to use the batch norm which is there in speech_cnn.

astorfi commented 6 years ago

The architecture is not designed to have good validation accuracy for the development dataset by simply using the softmax. It is only for creating the background model. Please refer to Issue #34

astorfi commented 6 years ago

@Abhishek12331 For the batch norm, the parameter is there and you can activate and deactivate it.

NightFury10497 commented 5 years ago

Hi,

I am getting a training accuracy of 95% on voxceleb but testing accuracy is around 10% only. What can be reason of this? Speakers=1211 Batch size =100 epoch = 50 Data size = 24000

can you tell how to integrated the training data on the code?