search

taylorlu / Speaker-Diarization

speaker diarization by uis-rnn and speaker embedding by vgg-speaker-recognition
Apache License 2.0
470 stars 121 forks source link

Where can I find ghostvlad/training_data.npz ? #51

Open yuanlorna opened 3 years ago

yuanlorna commented 3 years ago

May I ask how "./ghostvlad/training_data.npz" is generated, In addition, which dataset is the "pretrained/saved_model.uisrnn_benchmark" model trained on?

FlorentF9 commented 3 years ago

As explained in the readme, "This project only shows how to generate speaker embeddings using pre-trained model for uis-rnn training in later." I suppose information on how to train the speaker embeddings can be found in this repo or this one (official VGG code). It was trained on openslr, VCTK and VoxCeleb datasets.

Concerning the pretrained UISRNN weights, I assume it was trained on the 4 speakers dataset available in this repo, but it'd be great to have a clear confirmation.