bug in metric calculation

audio_toolset/audio_io.py

@@ -88,10 +88,7 @@ class LibrosaAudioToMel(object):
     def __init__(self, amplitude_to_db=False, power_to_db=False, **mel_kwargs):
         assert not all([amplitude_to_db, power_to_db]), "Choose amplitude_to_db or power_to_db, not both!"
         # Mel kwargs are:
-        #   sr
-        #   n_mels
-        #   n_fft
-        #   hop_length
+        #   sr   n_mels   n_fft   hop_length
 
         self.mel_kwargs = mel_kwargs
         self.amplitude_to_db = amplitude_to_db

audio_toolset/audio_to_mel_dataset.py

@@ -34,10 +34,10 @@ class LibrosaAudioToMelDataset(Dataset):
         self.audio_path = Path(audio_file_path)
 
         mel_folder_suffix = self.audio_path.parent.parent.name
+        self.mel_folder = Path(str(self.audio_path)
+                               .replace(mel_folder_suffix, f'{mel_folder_suffix}_mel_folder')).parent.parent
 
-        self.mel_file_path = Path(str(self.audio_path)
-                                  .replace(mel_folder_suffix, f'{mel_folder_suffix}_mel_folder')
-                                  .replace(self.audio_path.suffix, '.npy'))
+        self.mel_file_path = self.mel_folder / f'{self.audio_path.stem}.npy'
 
         self.audio_augmentations = audio_augmentations
 
@@ -45,7 +45,7 @@ class LibrosaAudioToMelDataset(Dataset):
                                        self.audio_file_duration, mel_kwargs['sr'], mel_kwargs['hop_length'],
                                        mel_kwargs['n_mels'], transform=mel_augmentations)
 
-        self._mel_transform = Compose([LibrosaAudioToMel(**mel_kwargs),
+        self._mel_transform = Compose([LibrosaAudioToMel(power_to_db=False, **mel_kwargs),
                                        MelToImage()
                                        ])
 

audio_toolset/mel_augmentation.py

@@ -64,9 +64,9 @@ class ShiftTime(_BaseTransformation):
             # Set to silence for heading/ tailing
             shift = int(shift)
             if shift > 0:
-                augmented_data[:shift, :] = 0
+                augmented_data[:, :shift] = 0
             else:
-                augmented_data[shift:, :] = 0
+                augmented_data[:, shift:] = 0
             return augmented_data
         else:
             return x