Netflix Mines Own Content to Create Dataset for Speech and Music Activity Detection
Identifying speech and music in audio files is a necessary step before dubbing, but open-source datasets fall short. Streaming giant Netflix built its own and says it outperforms.