Pitch and fargan model updates
Removing one of the 2d conv layers for pitch estimation reduces complexity without noticeable degradation. FARGAN model has more adversarial training. Also, no need for the double precision in the low-pass filter.
Showing
- autogen.sh 1 addition, 1 deletionautogen.sh
- dnn/lpcnet_enc.c 2 additions, 2 deletionsdnn/lpcnet_enc.c
- dnn/pitchdnn.c 1 addition, 2 deletionsdnn/pitchdnn.c
- dnn/torch/neural-pitch/export_neuralpitch_weights.py 1 addition, 2 deletionsdnn/torch/neural-pitch/export_neuralpitch_weights.py
- dnn/torch/neural-pitch/models.py 2 additions, 5 deletionsdnn/torch/neural-pitch/models.py
Loading
Please register or sign in to comment