Collection of different data selection methods, where the amount of training data is the same
Dragoș Alexandru Bălan
greenw0lf
AI & ML interests
Automatic speech recognition for low-resource languages
Organizations
None yet
Collection of different data selection methods, where the amount of training data is the same
Collection of Whisper FT models using different data selection approaches (metadata, classifier, full corpus)
Models that have all been trained with 30 hours of speech, but using different ratios of child-adult speech