YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
PN Incoder / Positive-Negative Enrollment Encoder
This repository contains the PN enrollment encoder and checkpoint used in the PN-conditioned FlowSE Target Speaker Extraction experiments.
Contents
checkpoints/proposed-monaural.pt- PN baseline / PN enrollment checkpoint.
- Used as the frozen PN enrollment encoder checkpoint.
model/- PN model implementation files.
dataset/- Dataset utilities used by the PN enrollment model.
Baseline Result
The PN baseline evaluation produced:
- Imp SNR: 10.65806 dB
- Imp SI-SNR: 8.76132 dB
- Imp SI-SDR: 9.38149 dB
Usage in PN-conditioned FlowSE
This checkpoint is used to extract positive-negative enrollment speaker condition embeddings.
The extracted condition embedding is then injected into the FlowSE DiT/CFM decoder for Target Speaker Extraction.
Note
The PN-conditioned FlowSE checkpoint is still being trained separately and is not included here.
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support