metadata
license: apache-2.0
Repository to store all the results and artifacts produced.
WIP content overview:
- audio_features: for each recording stores the audio features extract with Praat
- charts: charts produced in notebooks and somewhere else
- dataset_statistics: info about datasets and records. "support" files contain, for each row, the results of the voice activity detection pipeline.
- embeddings: "ecapa" embeddings come from the speaker verification pipeline and are used in the speaker id extraction pipeline for fleurs.