model performance using gendered audio recordings dataset for categorical_crossentropy metric.png