/deepspeaker-pytorch

Primary LanguagePython

deepspeaker-pytorch

  • pip install -r requirement.txt

*** Training

  • copy the data to your directory
  • change the line to point to your train directory
train_dir = DeepSpeakerDataset(path = YOUR_PATH_TO_TRAIN_DATA,n_triplets=args.n_triplets,loader = file_loader,transform=transform)
  • open the terminal and enter python3 train_triplet.py --epochs 1 for example to train one epoch.

*** Enrollment and Test

  • copy the test data to your directory
  • edit the path in main funcion of files.py to point to test data, and python3 files.py to generate file test.txt for each person
  • go to each person's directory and create enrollment.txt, manually copy the first 5 lines of test.txt to enrollment.txt and delete them from test.txt
  • edit _path in enrollment_test.py to point to test data
  • python3 enrollment_test.py --resume [PATH_TO_PTH_MODEL_FILE]