def __getitem__(self, index): example = self.collection[index] features = AudioSegment.segment_from_file( example.audio_file, n_segments=self.n_segments, trim=self.trim, ) features = torch.tensor(features.samples, dtype=torch.float) f, fl = features, torch.tensor(features.shape[0]).long() return f, fl
def __getitem__(self, index): """ Given a index, returns audio and audio_length of the corresponding element. Audio clips of n_segments are randomly chosen if the audio is longer than n_segments. """ example = self.collection[index] features = AudioSegment.segment_from_file(example.audio_file, n_segments=self.n_segments, trim=self.trim,) features = torch.tensor(features.samples) audio, audio_length = features, torch.tensor(features.shape[0]).long() truncate = audio_length % self.truncate_to if truncate != 0: audio_length -= truncate.long() audio = audio[:audio_length] return audio, audio_length