transition

_paramters.py

@@ -22,7 +22,7 @@ main_arg_parser.add_argument("--main_seed", type=int, default=69, help="")
 main_arg_parser.add_argument("--data_class_name", type=str, default='Urban8K', help="")
 main_arg_parser.add_argument("--data_worker", type=int, default=6, help="")
 main_arg_parser.add_argument("--data_root", type=str, default='data', help="")
-main_arg_parser.add_argument("--data_use_preprocessed", type=strtobool, default=True, help="")
+main_arg_parser.add_argument("--data_reset", type=strtobool, default=False, help="")
 main_arg_parser.add_argument("--data_n_mels", type=int, default=64, help="")
 main_arg_parser.add_argument("--data_sr", type=int, default=16000, help="")
 main_arg_parser.add_argument("--data_hop_length", type=int, default=256, help="")

datasets/base_dataset.py

@@ -0,0 +1,104 @@
+import pickle
+from pathlib import Path
+from typing import Union
+from abc import ABC
+
+import variables as V
+
+from torch.utils.data import Dataset
+
+class BaseAudioToMelDataset(Dataset, ABC):
+
+    @property
+    def task_type(self):
+        return self._task_type
+
+    @property
+    def classes(self):
+        return V.multi_classes
+
+    @property
+    def n_classes(self):
+        return V.N_CLASS_binary if self.task_type == V.TASK_OPTION_binary else V.N_CLASS_multi
+
+    @property
+    def sample_shape(self):
+        return self[0][0].shape
+
+    @property
+    def _fingerprint(self):
+        raise NotImplementedError
+        return str(self._mel_transform)
+
+    # Data Structures
+    @property
+    def mel_folder(self):
+        return self.data_root / 'mel'
+
+    @property
+    def wav_folder(self):
+        return self.data_root / self._wav_folder_name
+
+    @property
+    def _container_ext(self):
+        return '.mel'
+
+    def __init__(self, data_root: Union[str, Path], task_type, mel_kwargs,
+                 mel_augmentations=None, audio_augmentations=None, reset=False,
+                 wav_folder_name='wav', **_):
+        super(BaseAudioToMelDataset, self).__init__()
+
+        # Dataset Parameters
+        self.data_root = Path(data_root)
+        self._wav_folder_name = wav_folder_name
+        self.reset = reset
+        self.mel_kwargs = mel_kwargs
+
+        # Transformations
+        self.mel_augmentations = mel_augmentations
+        self.audio_augmentations = audio_augmentations
+        self._task_type = task_type
+
+        # Find all raw files and turn generator to persistent list:
+        self._wav_files = list(self.wav_folder.rglob('*.wav'))
+
+        # Build the Dataset
+        self._dataset = self._build_dataset()
+
+
+    def __len__(self):
+        raise NotImplementedError
+
+    def __getitem__(self, item):
+        raise NotImplementedError
+
+    def _build_dataset(self):
+        raise NotImplementedError
+
+    def _check_reset_and_clean_up(self, reset):
+        all_mel_folders = set([str(x.parent).replace(self._wav_folder_name, 'mel') for x in self._wav_files])
+        for mel_folder in all_mel_folders:
+            param_storage = Path(mel_folder) / 'data_params.pik'
+            param_storage.parent.mkdir(parents=True, exist_ok=True)
+            try:
+                pik_data = param_storage.read_bytes()
+                fingerprint = pickle.loads(pik_data)
+                if fingerprint == self._fingerprint:
+                    this_reset = reset
+                else:
+                    print('Diverging parameters were found; Refreshing...')
+                    param_storage.unlink()
+                    pik_data = pickle.dumps(self._fingerprint)
+                    param_storage.write_bytes(pik_data)
+                    this_reset = True
+
+            except FileNotFoundError:
+                pik_data = pickle.dumps(self._fingerprint)
+                param_storage.write_bytes(pik_data)
+                this_reset = True
+
+            if this_reset:
+                all_mel_files = self.mel_folder.rglob(f'*{self._container_ext}')
+                for mel_file in all_mel_files:
+                    mel_file.unlink()
+

datasets/binar_masks.py

@@ -20,17 +20,20 @@ class BinaryMasksDataset(Dataset):
 
     @property
     def _fingerprint(self):
-        return dict(**self._mel_kwargs, normalize=self.normalize)
+        return dict(**self._mel_kwargs if self._mel_kwargs else dict())
 
     def __init__(self, data_root, setting, mel_transforms, transforms=None, stretch_dataset=False,
-                 use_preprocessed=True):
+                 use_preprocessed=True, mel_kwargs=None):
         self.stretch = stretch_dataset
         assert isinstance(setting, str), f'Setting has to be a string, but was: {type(setting)}.'
         assert setting in V.DATA_OPTIONS, f'Setting must match one of: {V.DATA_OPTIONS}.'
         super(BinaryMasksDataset, self).__init__()
 
+        self.task = V.TASK_OPTION_binary
+
         self.data_root = Path(data_root) / 'ComParE2020_Mask'
         self.setting = setting
+        self._mel_kwargs = mel_kwargs
         self._wav_folder = self.data_root / 'wav'
         self._mel_folder = self.data_root / 'mel'
         self.container_ext = '.pik'

datasets/urban_8k.py

@@ -1,140 +1,78 @@
-import pickle
 from pathlib import Path
-import multiprocessing as mp
+from typing import Union, List
 
-import librosa as librosa
-from torch.utils.data import Dataset, ConcatDataset
+import multiprocessing as mp
+from torch.utils.data import ConcatDataset
 import torch
 from tqdm import tqdm
 
 import variables as V
-from ml_lib.audio_toolset.mel_dataset import TorchMelDataset
-from ml_lib.modules.util import F_x
+from datasets.base_dataset import BaseAudioToMelDataset
+from ml_lib.audio_toolset.audio_to_mel_dataset import LibrosaAudioToMelDataset, PyTorchAudioToMelDataset
 
 
-class Urban8K(Dataset):
+try:
+    torch.multiprocessing.set_sharing_strategy('file_system')
+except AttributeError:
+    pass
 
-    @property
-    def sample_shape(self):
-        return self[0][0].shape
+class Urban8K(BaseAudioToMelDataset):
 
-    @property
-    def _fingerprint(self):
-        return str(self._mel_transform)
-
-    def __init__(self, data_root, setting, mel_transforms, fold=1, transforms=None,
-                 use_preprocessed=True, audio_segment_len=62, audio_hop_len=30, num_worker=mp.cpu_count(),
-                 **_):
+    def __init__(self,
+                 data_root, setting, fold: Union[int, List]=1, num_worker=mp.cpu_count(),
+                 reset=False, sample_segment_len=50, sample_hop_len=20,
+                 **kwargs):
+        self.num_worker = num_worker
         assert isinstance(setting, str), f'Setting has to be a string, but was: {type(setting)}.'
         assert setting in V.DATA_OPTIONS, f'Setting must match one of: {V.DATA_OPTIONS}.'
-        assert fold in range(1, 11)
-        super(Urban8K, self).__init__()
+        assert fold in range(1, 11) if isinstance(fold, int) else all([f in range(1, 11) for f in fold])
 
-        self.data_root = Path(data_root) / 'UrbanSound8K'
+        #Dataset Paramters
         self.setting = setting
-        self.num_worker = num_worker
-        self.fold = fold if self.setting == V.DATA_OPTIONS.train else 10
-        self.use_preprocessed = use_preprocessed
-        self._wav_folder = self.data_root / 'audio' / f'fold{self.fold}'
-        self._mel_folder = self.data_root / 'mel' / f'fold{self.fold}'
-        self.container_ext = '.pik'
-        self._mel_transform = mel_transforms
+        fold = fold if self.setting != V.DATA_OPTION_test else 10
+        self.fold = fold if isinstance(fold, list) else [fold]
 
-        self._labels = self._build_labels()
-        self._wav_files = list(sorted(self._labels.keys()))
-        transforms = transforms or F_x(in_shape=None)
+        self.sample_segment_len = sample_segment_len
+        self.sample_hop_len = sample_hop_len
 
-        param_storage = self._mel_folder / 'data_params.pik'
-        self._mel_folder.mkdir(parents=True, exist_ok=True)
-        try:
-            pik_data = param_storage.read_bytes()
-            fingerprint = pickle.loads(pik_data)
-            if fingerprint == self._fingerprint:
-                self.use_preprocessed = use_preprocessed
-            else:
-                print('Diverging parameters were found; Refreshing...')
-                param_storage.unlink()
-                pik_data = pickle.dumps(self._fingerprint)
-                param_storage.write_bytes(pik_data)
-                self.use_preprocessed = False
+        # Dataset specific super init
+        super(Urban8K, self).__init__(Path(data_root) / 'UrbanSound8K',
+                                      V.TASK_OPTION_multiclass, reset=reset, wav_folder_name='audio', **kwargs
+                                      )
 
-        except FileNotFoundError:
-            pik_data = pickle.dumps(self._fingerprint)
-            param_storage.write_bytes(pik_data)
-            self.use_preprocessed = False
+    def _build_subdataset(self, row):
+        slice_file_name, fs_id, start, end, salience, fold, class_id, class_name = row.strip().split(',')
+        fold, class_id = (int(x) for x in (fold, class_id))
+        if int(fold) in self.fold:
+            audio_file_path = self.wav_folder / f'fold{fold}' / slice_file_name
+            return PyTorchAudioToMelDataset(audio_file_path, class_id, **self.__dict__)
+        else:
+            return None
 
-
-        while True:
-            if not self.use_preprocessed:
-                self._pre_process()
-            try:
-                self._dataset = ConcatDataset(
-                    [TorchMelDataset(identifier=key, mel_path=self._mel_folder, transform=transforms,
-                                     segment_len=audio_segment_len, hop_len=audio_hop_len,
-                                     label=self._labels[key]['label']
-                                     ) for key in self._labels.keys()]
-                )
-                break
-            except IOError:
-                self.use_preprocessed = False
-                pass
-
-    def _build_labels(self):
-        labeldict = dict()
+    def _build_dataset(self):
+        dataset= list()
         with open(Path(self.data_root) / 'metadata' / 'UrbanSound8K.csv', mode='r') as f:
             # Exclude the header
             _ = next(f)
-            for row in f:
-                slice_file_name, fs_id, start, end, salience, fold, class_id, class_name = row.strip().split(',')
-                if int(fold) == self.fold:
-                    key = slice_file_name.replace('.wav', '')
-                    labeldict[key] = dict(label=int(class_id), fold=int(fold))
+            all_rows = list(f)
+            chunksize = len(all_rows) // max(self.num_worker, 1)
+            with mp.Pool(processes=self.num_worker) as pool:
+                with tqdm(total=len(all_rows)) as pbar:
+                    for i, sub_dataset in enumerate(
+                            pool.imap_unordered(self._build_subdataset, all_rows, chunksize=chunksize)):
+                        pbar.update()
+                        dataset.append(sub_dataset)
 
-        # Delete File if one exists.
-        if not self.use_preprocessed:
-            for key in labeldict.keys():
-                for mel_file in self._mel_folder.rglob(f'{key}_*'):
-                    try:
-                        mel_file.unlink(missing_ok=True)
-                    except FileNotFoundError:
-                        pass
-
-        return labeldict
+        dataset = ConcatDataset([x for x in dataset if x is not None])
+        return dataset
 
     def __len__(self):
         return len(self._dataset)
 
-    def _pre_process(self):
-        print('Preprocessing Mel Files....')
-        with mp.Pool(processes=self.num_worker) as pool:
-            with tqdm(total=len(self._labels)) as pbar:
-                for i, _ in enumerate(pool.imap_unordered(self._build_mel,  self._labels.keys())):
-                    pbar.update()
-
-    def _build_mel(self, filename):
-
-        wav_file = self._wav_folder / (filename.replace('X', '') + '.wav')
-        mel_file = list(self._mel_folder.glob(f'{filename}_*'))
-
-        if not mel_file:
-            raw_sample, sr = librosa.core.load(wav_file)
-            mel_sample = self._mel_transform(raw_sample)
-            m, n = mel_sample.shape
-            mel_file = self._mel_folder / f'{filename}_{m}_{n}'
-            self._mel_folder.mkdir(exist_ok=True, parents=True)
-            with mel_file.open(mode='wb') as f:
-                pickle.dump(mel_sample, f, protocol=pickle.HIGHEST_PROTOCOL)
-        else:
-            # print(f"Already existed.. Skipping {filename}")
-            mel_file = mel_file[0]
-
-        with mel_file.open(mode='rb') as f:
-            mel_sample = pickle.load(f, fix_imports=True)
-        return mel_sample, mel_file
 
     def __getitem__(self, item):
         transformed_samples, label = self._dataset[item]
 
-        label = torch.as_tensor(label, dtype=torch.float)
+        label = torch.as_tensor(label, dtype=torch.int)
 
         return transformed_samples, label

datasets/urban_8k_torchaudio.py

@@ -1,140 +0,0 @@
-import pickle
-from pathlib import Path
-import multiprocessing as mp
-
-import librosa as librosa
-from torch.utils.data import Dataset, ConcatDataset
-import torch
-from tqdm import tqdm
-
-import variables as V
-from ml_lib.audio_toolset.mel_dataset import TorchMelDataset
-from ml_lib.modules.util import F_x
-
-
-class Urban8K_TO(Dataset):
-
-    @property
-    def sample_shape(self):
-        return self[0][0].shape
-
-    @property
-    def _fingerprint(self):
-        return str(self._mel_transform)
-
-    def __init__(self, data_root, setting, mel_transforms, fold=1, transforms=None,
-                 use_preprocessed=True, audio_segment_len=1, audio_hop_len=1, num_worker=mp.cpu_count(),
-                 **_):
-        assert isinstance(setting, str), f'Setting has to be a string, but was: {type(setting)}.'
-        assert setting in V.DATA_OPTIONS, f'Setting must match one of: {V.DATA_OPTIONS}.'
-        assert fold in range(1, 11)
-        super(Urban8K_TO, self).__init__()
-
-        self.data_root = Path(data_root) / 'UrbanSound8K'
-        self.setting = setting
-        self.num_worker = num_worker
-        self.fold = fold if self.setting == V.DATA_OPTIONS.train else 10
-        self.use_preprocessed = use_preprocessed
-        self._wav_folder = self.data_root / 'audio' / f'fold{self.fold}'
-        self._mel_folder = self.data_root / 'mel' / f'fold{self.fold}'
-        self.container_ext = '.pik'
-        self._mel_transform = mel_transforms
-
-        self._labels = self._build_labels()
-        self._wav_files = list(sorted(self._labels.keys()))
-        transforms = transforms or F_x(in_shape=None)
-
-        param_storage = self._mel_folder / 'data_params.pik'
-        self._mel_folder.mkdir(parents=True, exist_ok=True)
-        try:
-            pik_data = param_storage.read_bytes()
-            fingerprint = pickle.loads(pik_data)
-            if fingerprint == self._fingerprint:
-                self.use_preprocessed = use_preprocessed
-            else:
-                print('Diverging parameters were found; Refreshing...')
-                param_storage.unlink()
-                pik_data = pickle.dumps(self._fingerprint)
-                param_storage.write_bytes(pik_data)
-                self.use_preprocessed = False
-
-        except FileNotFoundError:
-            pik_data = pickle.dumps(self._fingerprint)
-            param_storage.write_bytes(pik_data)
-            self.use_preprocessed = False
-
-
-        while True:
-            if not self.use_preprocessed:
-                self._pre_process()
-            try:
-                self._dataset = ConcatDataset(
-                    [TorchMelDataset(identifier=key, mel_path=self._mel_folder, transform=transforms,
-                                     segment_len=audio_segment_len, hop_len=audio_hop_len,
-                                     label=self._labels[key]['label']
-                                     ) for key in self._labels.keys()]
-                )
-                break
-            except IOError:
-                self.use_preprocessed = False
-                pass
-
-    def _build_labels(self):
-        labeldict = dict()
-        with open(Path(self.data_root) / 'metadata' / 'UrbanSound8K.csv', mode='r') as f:
-            # Exclude the header
-            _ = next(f)
-            for row in f:
-                slice_file_name, fs_id, start, end, salience, fold, class_id, class_name = row.strip().split(',')
-                if int(fold) == self.fold:
-                    key = slice_file_name.replace('.wav', '')
-                    labeldict[key] = dict(label=int(class_id), fold=int(fold))
-
-        # Delete File if one exists.
-        if not self.use_preprocessed:
-            for key in labeldict.keys():
-                for mel_file in self._mel_folder.rglob(f'{key}_*'):
-                    try:
-                        mel_file.unlink(missing_ok=True)
-                    except FileNotFoundError:
-                        pass
-
-        return labeldict
-
-    def __len__(self):
-        return len(self._dataset)
-
-    def _pre_process(self):
-        print('Preprocessing Mel Files....')
-        with mp.Pool(processes=self.num_worker) as pool:
-            with tqdm(total=len(self._labels)) as pbar:
-                for i, _ in enumerate(pool.imap_unordered(self._build_mel,  self._labels.keys())):
-                    pbar.update()
-
-    def _build_mel(self, filename):
-
-        wav_file = self._wav_folder / (filename.replace('X', '') + '.wav')
-        mel_file = list(self._mel_folder.glob(f'{filename}_*'))
-
-        if not mel_file:
-            raw_sample, sr = librosa.core.load(wav_file)
-            mel_sample = self._mel_transform(raw_sample)
-            m, n = mel_sample.shape
-            mel_file = self._mel_folder / f'{filename}_{m}_{n}'
-            self._mel_folder.mkdir(exist_ok=True, parents=True)
-            with mel_file.open(mode='wb') as f:
-                pickle.dump(mel_sample, f, protocol=pickle.HIGHEST_PROTOCOL)
-        else:
-            # print(f"Already existed.. Skipping {filename}")
-            mel_file = mel_file[0]
-
-        with mel_file.open(mode='rb') as f:
-            mel_sample = pickle.load(f, fix_imports=True)
-        return mel_sample, mel_file
-
-    def __getitem__(self, item):
-        transformed_samples, label = self._dataset[item]
-
-        label = torch.as_tensor(label, dtype=torch.float)
-
-        return transformed_samples, label

ensemble_methods/global_inference.py

@@ -9,7 +9,7 @@ from torch.utils.data import DataLoader, Dataset
 from torchvision.transforms import Compose, RandomApply
 
 from ml_lib.audio_toolset.audio_augmentation import Speed
-from ml_lib.audio_toolset.audio_io import AudioToMel, NormalizeLocal, MelToImage
+from ml_lib.audio_toolset.audio_io import LibrosaAudioToMel, NormalizeLocal, MelToImage
 
 # Dataset and Dataloaders
 # =============================================================================
@@ -28,8 +28,8 @@ from datasets.binar_masks import BinaryMasksDataset
 
 def prepare_dataloader(config_obj):
     mel_transforms = Compose([
-        AudioToMel(sr=config_obj.data.sr, n_mels=config_obj.data.n_mels, n_fft=config_obj.data.n_fft,
-                   hop_length=config_obj.data.hop_length),
+        LibrosaAudioToMel(sr=config_obj.data.sr, n_mels=config_obj.data.n_mels, n_fft=config_obj.data.n_fft,
+                          hop_length=config_obj.data.hop_length),
         MelToImage()])
     transforms = Compose([NormalizeLocal(), ToTensor()])
     """    

main_inference.py

@@ -8,7 +8,7 @@ from torch.utils.data import DataLoader, Dataset
 from torchvision.transforms import Compose, RandomApply
 
 from ml_lib.audio_toolset.audio_augmentation import Speed
-from ml_lib.audio_toolset.audio_io import AudioToMel, NormalizeLocal, MelToImage
+from ml_lib.audio_toolset.audio_io import LibrosaAudioToMel, NormalizeLocal, MelToImage
 
 # Dataset and Dataloaders
 # =============================================================================
@@ -26,8 +26,8 @@ from datasets.binar_masks import BinaryMasksDataset
 
 def prepare_dataloader(config_obj):
     mel_transforms = Compose([
-        AudioToMel(sr=config_obj.data.sr, n_mels=config_obj.data.n_mels, n_fft=config_obj.data.n_fft,
-                   hop_length=config_obj.data.hop_length),
+        LibrosaAudioToMel(sr=config_obj.data.sr, n_mels=config_obj.data.n_mels, n_fft=config_obj.data.n_fft,
+                          hop_length=config_obj.data.hop_length),
         MelToImage()])
     transforms = Compose([NormalizeLocal(), ToTensor()])
     aug_transforms = Compose([

models/transformer_model.py

@@ -1,114 +0,0 @@
-from argparse import Namespace
-
-import warnings
-
-import torch
-from torch import nn
-
-from einops import rearrange, repeat
-
-from ml_lib.modules.blocks import TransformerModule
-from ml_lib.modules.util import (LightningBaseModule, AutoPadToShape, F_x)
-from util.module_mixins import (BaseOptimizerMixin, BaseTrainMixin, BaseValMixin, DatasetMixin,
-                                BaseDataloadersMixin, BaseTestMixin)
-
-MIN_NUM_PATCHES = 16
-
-class VisualTransformer(DatasetMixin,
-                        BaseDataloadersMixin,
-                        BaseTrainMixin,
-                        BaseValMixin,
-                        BaseTestMixin,
-                        BaseOptimizerMixin,
-                        LightningBaseModule
-                        ):
-
-    def __init__(self, hparams):
-        super(VisualTransformer, self).__init__(hparams)
-
-        # Dataset
-        # =============================================================================
-        self.dataset = self.build_dataset()
-
-        self.in_shape = self.dataset.train_dataset.sample_shape
-        assert len(self.in_shape) == 3, 'There need to be three Dimensions'
-        channels, height, width = self.in_shape
-
-        # Model Paramters
-        # =============================================================================
-        # Additional parameters
-        self.embed_dim = self.params.embedding_size
-
-        # Automatic Image Shaping
-        self.patch_size = self.params.patch_size
-        image_size = (max(height, width) // self.patch_size) * self.patch_size
-        self.image_size = image_size + self.patch_size if image_size < max(height, width) else image_size
-
-        # This should be obsolete
-        assert self.image_size % self.patch_size == 0, 'image dimensions must be divisible by the patch size'
-
-        num_patches = (self.image_size // self.patch_size) ** 2
-        patch_dim = channels * self.patch_size ** 2
-        assert num_patches >= MIN_NUM_PATCHES, f'your number of patches ({num_patches}) is way too small for ' + \
-                                               f'attention. Try decreasing your patch size'
-
-        # Correct the Embedding Dim
-        if not self.embed_dim % self.params.heads == 0:
-            self.embed_dim = (self.embed_dim // self.params.heads) * self.params.heads
-            message = ('Embedding Dimension was fixed to be devideable by the number' +
-                       f' of attention heads, is now: {self.embed_dim}')
-            for func in print, warnings.warn:
-                func(message)
-
-        # Utility Modules
-        self.autopad = AutoPadToShape((self.image_size, self.image_size))
-
-        # Modules with Parameters
-        self.transformer = TransformerModule(in_shape=self.embed_dim, hidden_size=self.params.lat_dim,
-                                             n_heads=self.params.heads, num_layers=self.params.attn_depth,
-                                             dropout=self.params.dropout, use_norm=self.params.use_norm,
-                                             activation=self.params.activation_as_string
-                                             )
-
-
-        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, self.embed_dim))
-        self.patch_to_embedding = nn.Linear(patch_dim, self.embed_dim) if self.params.embedding_size \
-            else F_x(self.embed_dim)
-        self.cls_token = nn.Parameter(torch.randn(1, 1, self.embed_dim))
-        self.dropout = nn.Dropout(self.params.dropout)
-
-        self.to_cls_token = nn.Identity()
-
-        self.mlp_head = nn.Sequential(
-            nn.LayerNorm(self.embed_dim),
-            nn.Linear(self.embed_dim, self.params.lat_dim),
-            nn.GELU(),
-            nn.Dropout(self.params.dropout),
-            nn.Linear(self.params.lat_dim, 1),
-            nn.Sigmoid()
-        )
-
-    def forward(self, x, mask=None):
-        """
-        :param x: the sequence to the encoder (required).
-        :param mask: the mask for the src sequence (optional).
-        :return:
-        """
-        tensor = self.autopad(x)
-        p = self.params.patch_size
-        tensor = rearrange(tensor, 'b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = p, p2 = p)
-
-        tensor = self.patch_to_embedding(tensor)
-        b, n, _ = tensor.shape
-
-        cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
-
-        tensor = torch.cat((cls_tokens, tensor), dim=1)
-        tensor += self.pos_embedding[:, :(n + 1)]
-        tensor = self.dropout(tensor)
-
-        tensor = self.transformer(tensor, mask)
-
-        tensor = self.to_cls_token(tensor[:, 0])
-        tensor = self.mlp_head(tensor)
-        return Namespace(main_out=tensor)

models/transformer_model_horizontal.py

@@ -1,111 +0,0 @@
-from argparse import Namespace
-
-import warnings
-
-import torch
-from torch import nn
-
-from ml_lib.modules.blocks import TransformerModule
-from ml_lib.modules.util import (LightningBaseModule, AutoPadToShape, F_x, SlidingWindow)
-from util.module_mixins import (BaseOptimizerMixin, BaseTrainMixin, BaseValMixin, DatasetMixin,
-                                BaseDataloadersMixin, BaseTestMixin)
-
-MIN_NUM_PATCHES = 16
-
-class HorizontalVisualTransformer(DatasetMixin,
-                                  BaseDataloadersMixin,
-                                  BaseTrainMixin,
-                                  BaseValMixin,
-                                  BaseTestMixin,
-                                  BaseOptimizerMixin,
-                                  LightningBaseModule
-                                  ):
-
-    def __init__(self, hparams):
-        super(HorizontalVisualTransformer, self).__init__(hparams)
-
-        # Dataset
-        # =============================================================================
-        self.dataset = self.build_dataset()
-
-        self.in_shape = self.dataset.train_dataset.sample_shape
-        assert len(self.in_shape) == 3, 'There need to be three Dimensions'
-        channels, height, width = self.in_shape
-
-        # Model Paramters
-        # =============================================================================
-        # Additional parameters
-        self.embed_dim = self.params.embedding_size
-        self.patch_size = self.params.patch_size
-        self.height = height
-        self.width = width
-        self.channels = channels
-
-        self.new_height = ((self.height - self.patch_size)//1) + 1
-
-        num_patches = self.new_height - (self.patch_size // 2)
-        patch_dim = channels * self.patch_size * self.width
-        assert num_patches >= MIN_NUM_PATCHES, f'your number of patches ({num_patches}) is way too small for ' + \
-                                               f'attention. Try decreasing your patch size'
-
-        # Correct the Embedding Dim
-        if not self.embed_dim % self.params.heads == 0:
-            self.embed_dim = (self.embed_dim // self.params.heads) * self.params.heads
-            message = ('Embedding Dimension was fixed to be devideable by the number' +
-                       f' of attention heads, is now: {self.embed_dim}')
-            for func in print, warnings.warn:
-                func(message)
-
-        # Utility Modules
-        self.autopad = AutoPadToShape((self.new_height, self.width))
-        self.dropout = nn.Dropout(self.params.dropout)
-        self.slider = SlidingWindow((channels, *self.autopad.target_shape), (self.patch_size, self.width),
-                                    keepdim=False)
-
-        # Modules with Parameters
-        self.transformer = TransformerModule(in_shape=self.embed_dim, hidden_size=self.params.lat_dim,
-                                             n_heads=self.params.heads, num_layers=self.params.attn_depth,
-                                             dropout=self.params.dropout, use_norm=self.params.use_norm,
-                                             activation=self.params.activation_as_string
-                                             )
-
-
-        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, self.embed_dim))
-        self.patch_to_embedding = nn.Linear(patch_dim, self.embed_dim) if self.params.embedding_size \
-            else F_x(self.embed_dim)
-        self.cls_token = nn.Parameter(torch.randn(1, 1, self.embed_dim))
-        self.to_cls_token = nn.Identity()
-
-        self.mlp_head = nn.Sequential(
-            nn.LayerNorm(self.embed_dim),
-            nn.Linear(self.embed_dim, self.params.lat_dim),
-            nn.GELU(),
-            nn.Dropout(self.params.dropout),
-            nn.Linear(self.params.lat_dim, 1),
-            nn.Sigmoid()
-        )
-
-    def forward(self, x, mask=None):
-        """
-        :param x: the sequence to the encoder (required).
-        :param mask: the mask for the src sequence (optional).
-        :return:
-        """
-        tensor = self.autopad(x)
-        tensor = self.slider(tensor)
-
-        tensor = self.patch_to_embedding(tensor)
-        b, n, _ = tensor.shape
-
-        # cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
-        cls_tokens = self.cls_token.repeat((b, 1, 1))
-
-        tensor = torch.cat((cls_tokens, tensor), dim=1)
-        tensor += self.pos_embedding[:, :(n + 1)]
-        tensor = self.dropout(tensor)
-
-        tensor = self.transformer(tensor, mask)
-
-        tensor = self.to_cls_token(tensor[:, 0])
-        tensor = self.mlp_head(tensor)
-        return Namespace(main_out=tensor)

models/transformer_model_vertical.py

@@ -1,110 +0,0 @@
-from argparse import Namespace
-
-import warnings
-
-import torch
-from torch import nn
-
-from ml_lib.modules.blocks import TransformerModule
-from ml_lib.modules.util import (LightningBaseModule, AutoPadToShape, F_x, SlidingWindow)
-from util.module_mixins import (BaseOptimizerMixin, BaseTrainMixin, BaseValMixin, DatasetMixin,
-                                BaseDataloadersMixin, BaseTestMixin)
-
-MIN_NUM_PATCHES = 16
-
-class VerticalVisualTransformer(DatasetMixin,
-                                BaseDataloadersMixin,
-                                BaseTrainMixin,
-                                BaseValMixin,
-                                BaseTestMixin,
-                                BaseOptimizerMixin,
-                                LightningBaseModule
-                                ):
-
-    def __init__(self, hparams):
-        super(VerticalVisualTransformer, self).__init__(hparams)
-
-        # Dataset
-        # =============================================================================
-        self.dataset = self.build_dataset()
-
-        self.in_shape = self.dataset.train_dataset.sample_shape
-        assert len(self.in_shape) == 3, 'There need to be three Dimensions'
-        channels, height, width = self.in_shape
-
-        # Model Paramters
-        # =============================================================================
-        # Additional parameters
-        self.embed_dim = self.params.embedding_size
-        self.patch_size = self.params.patch_size
-        self.height = height
-        self.width = width
-        self.channels = channels
-
-        self.new_width = ((self.width - self.patch_size)//1) + 1
-
-        num_patches = self.new_width - (self.patch_size // 2)
-        patch_dim = channels * self.patch_size * self.height
-        assert num_patches >= MIN_NUM_PATCHES, f'your number of patches ({num_patches}) is way too small for ' + \
-                                               f'attention. Try decreasing your patch size'
-
-        # Correct the Embedding Dim
-        if not self.embed_dim % self.params.heads == 0:
-            self.embed_dim = (self.embed_dim // self.params.heads) * self.params.heads
-            message = ('Embedding Dimension was fixed to be devideable by the number' +
-                       f' of attention heads, is now: {self.embed_dim}')
-            for func in print, warnings.warn:
-                func(message)
-
-        # Utility Modules
-        self.autopad = AutoPadToShape((self.height, self.new_width))
-        self.dropout = nn.Dropout(self.params.dropout)
-        self.slider = SlidingWindow((channels, *self.autopad.target_shape), (self.height, self.patch_size), keepdim=False)
-
-        # Modules with Parameters
-        self.transformer = TransformerModule(in_shape=self.embed_dim, hidden_size=self.params.lat_dim,
-                                             n_heads=self.params.heads, num_layers=self.params.attn_depth,
-                                             dropout=self.params.dropout, use_norm=self.params.use_norm,
-                                             activation=self.params.activation_as_string
-                                             )
-
-
-        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, self.embed_dim))
-        self.patch_to_embedding = nn.Linear(patch_dim, self.embed_dim) if self.params.embedding_size \
-            else F_x(self.embed_dim)
-        self.cls_token = nn.Parameter(torch.randn(1, 1, self.embed_dim))
-        self.to_cls_token = nn.Identity()
-
-        self.mlp_head = nn.Sequential(
-            nn.LayerNorm(self.embed_dim),
-            nn.Linear(self.embed_dim, self.params.lat_dim),
-            nn.GELU(),
-            nn.Dropout(self.params.dropout),
-            nn.Linear(self.params.lat_dim, 1),
-            nn.Sigmoid()
-        )
-
-    def forward(self, x, mask=None):
-        """
-        :param x: the sequence to the encoder (required).
-        :param mask: the mask for the src sequence (optional).
-        :return:
-        """
-        tensor = self.autopad(x)
-        tensor = self.slider(tensor)
-
-        tensor = self.patch_to_embedding(tensor)
-        b, n, _ = tensor.shape
-
-        # cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
-        cls_tokens = self.cls_token.repeat((b, 1, 1))
-
-        tensor = torch.cat((cls_tokens, tensor), dim=1)
-        tensor += self.pos_embedding[:, :(n + 1)]
-        tensor = self.dropout(tensor)
-
-        tensor = self.transformer(tensor, mask)
-
-        tensor = self.to_cls_token(tensor[:, 0])
-        tensor = self.mlp_head(tensor)
-        return Namespace(main_out=tensor)

multi_run.py

@@ -14,64 +14,134 @@ warnings.filterwarnings('ignore', category=UserWarning)
 
 if __name__ == '__main__':
 
-    args = main_arg_parser.parse_args()
-    # Model Settings
-    config = Config().read_namespace(args)
+    if False:
+        args = main_arg_parser.parse_args()
+        # Model Settings
+        config = Config().read_namespace(args)
 
-    arg_dict = dict()
-    for seed in range(1):
-        arg_dict.update(main_seed=seed)
-        if False:
-            for patch_size in [3, 5 , 9]:
-                for model in ['VerticalVisualTransformer']:
-                    arg_dict.update(model_type=model, model_patch_size=patch_size)
-                    raw_conf = dict(data_speed_amount=0.0, data_speed_min=0.0, data_speed_max=0.0,
-                                    data_mask_ratio=0.0, data_noise_ratio=0.0, data_shift_ratio=0.0, data_loudness_ratio=0.0,
-                                    data_stretch=False, train_epochs=401)
+        arg_dict = dict()
+        for seed in range(1):
+            arg_dict.update(main_seed=seed)
+            if False:
+                for patch_size in [3, 5 , 9]:
+                    for model in ['VerticalVisualTransformer']:
+                        arg_dict.update(model_type=model, model_patch_size=patch_size)
+                        raw_conf = dict(data_speed_amount=0.0, data_speed_min=0.0, data_speed_max=0.0,
+                                        data_mask_ratio=0.0, data_noise_ratio=0.0, data_shift_ratio=0.0, data_loudness_ratio=0.0,
+                                        data_stretch=False, train_epochs=401)
 
-                    all_conf = dict(data_speed_amount=0.4, data_speed_min=0.7, data_speed_max=1.7,
-                                    data_mask_ratio=0.2, data_noise_ratio=0.4, data_shift_ratio=0.4, data_loudness_ratio=0.4,
-                                    data_stretch=True, train_epochs=101)
+                        all_conf = dict(data_speed_amount=0.4, data_speed_min=0.7, data_speed_max=1.7,
+                                        data_mask_ratio=0.2, data_noise_ratio=0.4, data_shift_ratio=0.4, data_loudness_ratio=0.4,
+                                        data_stretch=True, train_epochs=101)
 
-                    speed_conf = raw_conf.copy()
-                    speed_conf.update(data_speed_amount=0.4, data_speed_min=0.7, data_speed_max=1.7,
-                                      data_stretch=True, train_epochs=101)
+                        speed_conf = raw_conf.copy()
+                        speed_conf.update(data_speed_amount=0.4, data_speed_min=0.7, data_speed_max=1.7,
+                                          data_stretch=True, train_epochs=101)
 
-                    mask_conf = raw_conf.copy()
-                    mask_conf.update(data_mask_ratio=0.2, data_stretch=True, train_epochs=101)
+                        mask_conf = raw_conf.copy()
+                        mask_conf.update(data_mask_ratio=0.2, data_stretch=True, train_epochs=101)
 
-                    noise_conf = raw_conf.copy()
-                    noise_conf.update(data_noise_ratio=0.4, data_stretch=True, train_epochs=101)
+                        noise_conf = raw_conf.copy()
+                        noise_conf.update(data_noise_ratio=0.4, data_stretch=True, train_epochs=101)
 
-                    shift_conf = raw_conf.copy()
-                    shift_conf.update(data_shift_ratio=0.4, data_stretch=True, train_epochs=101)
+                        shift_conf = raw_conf.copy()
+                        shift_conf.update(data_shift_ratio=0.4, data_stretch=True, train_epochs=101)
 
-                    loudness_conf = raw_conf.copy()
-                    loudness_conf.update(data_loudness_ratio=0.4, data_stretch=True, train_epochs=101)
+                        loudness_conf = raw_conf.copy()
+                        loudness_conf.update(data_loudness_ratio=0.4, data_stretch=True, train_epochs=101)
 
-                    for dicts in [raw_conf, all_conf, speed_conf, mask_conf, noise_conf, shift_conf, loudness_conf]:
+                        for dicts in [raw_conf, all_conf, speed_conf, mask_conf, noise_conf, shift_conf, loudness_conf]:
+
+                            arg_dict.update(dicts)
+            if True:
+                for patch_size in [7]:
+                    for lat_dim in [32]:
+                        for heads in [8]:
+                            for embedding_size in [7**2]:
+                                for attn_depth in [1, 3, 5, 7]:
+                                    for model in ['HorizontalVisualTransformer']:
+                                        arg_dict.update(
+                                            model_type=model,
+                                            model_patch_size=patch_size,
+                                            model_lat_dim=lat_dim,
+                                            model_heads=heads,
+                                            model_embedding_size=embedding_size,
+                                            model_attn_depth=attn_depth
+                                        )
+                                        config = config.update(arg_dict)
+                                        version_path = config.exp_path / config.version
+                                        if version_path.exists():
+                                            if not (version_path / 'weights.ckpt').exists():
+                                                shutil.rmtree(version_path)
+                                            else:
+                                                continue
+                                        run_lightning_loop(config)
+
+    import matplotlib
+    import matplotlib.pyplot as plt
+    import matplotlib.cm as cm
+    import numpy as np
+
+    from diffractio import degrees, mm, plt, sp, um, np
+    from diffractio.scalar_fields_XY import Scalar_field_XY
+    from diffractio.utils_drawing import draw_several_fields
+    from diffractio.scalar_masks_XY import Scalar_mask_XY
+    from diffractio.scalar_sources_XY import Scalar_source_XY
+
+    from matplotlib import rcParams
+
+    rcParams['figure.figsize']=(7,5)
+    rcParams['figure.dpi']=75
+
+    period = 20 * um
+    num_pixels = 512
+
+    length = 250 * um
+    x0 = np.linspace(-length / 2, length / 2, num_pixels)
+    y0 = np.linspace(-length / 2, length / 2, num_pixels)
+    wavelength = 0.6238 * um
+
+    u1 = Scalar_source_XY(x=x0, y=y0, wavelength=wavelength)
+    u1.plane_wave(A=1, theta=0 * degrees, phi=0 * degrees)
+
+    t1 = Scalar_mask_XY(x=x0, y=y0, wavelength=wavelength)
+    t1.forked_grating(kind='amplitude',
+                      r0=(0 * um, 0 * um), period=period, l=3, alpha=2, angle=0 * degrees)
+
+    u2 = u1 * t1
+
+    t2 = Scalar_mask_XY(x=x0, y=y0, wavelength=wavelength)
+    t2.roughness(t=(20 * um, 20 * um), s=1 * um)
+
+    u2 = u2 * t2
+    u2.draw(kind='phase')
+
+    u3 = u2.RS(z=1 * mm, new_field=True)
+
+    u4 = u2.RS(z=5 * mm, new_field=True)
+
+    u5 = u2.RS(z=10 * mm, new_field=True)
+
+    print('draw')
+
+    draw_several_fields((u3, u4, u5), titulos=('1 mm', '5 mm', '10 mm'), logarithm=True)
+
+    plt.show()
+
+    pass
+
+    u2 = t2 * u1
+    u2.draw(kind='phase')
+
+    u3 = u2.RS(z=1 * mm, new_field=True)
+
+    u4 = u2.RS(z=5 * mm, new_field=True)
+
+    u5 = u2.RS(z=10 * mm, new_field=True)
+
+    print('draw')
+
+    draw_several_fields((u3, u4, u5), titulos=('1 mm', '5 mm', '10 mm'), logarithm=True)
+
+    plt.show()
 
-                        arg_dict.update(dicts)
-        if True:
-            for patch_size in [7]:
-                for lat_dim in [32]:
-                    for heads in [8]:
-                        for embedding_size in [7**2]:
-                            for attn_depth in [1, 3, 5, 7]:
-                                for model in ['HorizontalVisualTransformer']:
-                                    arg_dict.update(
-                                        model_type=model,
-                                        model_patch_size=patch_size,
-                                        model_lat_dim=lat_dim,
-                                        model_heads=heads,
-                                        model_embedding_size=embedding_size,
-                                        model_attn_depth=attn_depth
-                                    )
-                                    config = config.update(arg_dict)
-                                    version_path = config.exp_path / config.version
-                                    if version_path.exists():
-                                        if not (version_path / 'weights.ckpt').exists():
-                                            shutil.rmtree(version_path)
-                                        else:
-                                            continue
-                                    run_lightning_loop(config)

output/metrics.csv

@@ -0,0 +1,16 @@
+worker,root,class_name,normalized,use_preprocessed,n_mels,sr,hop_length,n_fft,mixup,stretch,loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_ratio,speed_factor,mean,max,median,std
+11,data,BinaryMasksDataset,True,False,64,16000,256,512,False,True,0,0,0,0,0,0,63.79,63.97,63.85,0.19
+11,data,BinaryMasksDataset,True,False,64,16000,256,512,False,True,0,0,0,0,0,0,64.2,64.9,64.02,0.5
+11,data,BinaryMasksDataset,True,False,64,16000,256,512,False,True,0,0,0,0,0,0,63.34,63.92,63.45,0.49
+11,data,BinaryMasksDataset,True,False,64,16000,256,512,False,True,0,0,0,0,0,0,63.92,64.34,63.91,0.25
+11,data,BinaryMasksDataset,True,False,64,16000,256,512,False,True,0,0,0,0,0,0,64.72,65.38,64.6,0.51
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,False,0,0,0,0,0,0,60.36,64.16,63.41,5.31
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,False,0,0,0,0,0,0,58.4,64.18,63.86,6.85
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,False,0,0,0,0,0,0,60.09,63.61,63.26,5.19
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,False,0,0,0,0,0,0,59.96,64.34,63.32,5.4
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,False,0,0,0,0,0,0,60.94,64.59,64.11,5.56
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,0,0,0.3,0,0.3,0.7,50.0,50.0,50.0,0.0
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,0,0,0.3,0,0.3,0.7,62.1,65.51,64.9,5.48
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,0,0,0.3,0,0.3,0.7,61.29,64.76,64.31,5.68
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,0,0,0.3,0,0.3,0.7,63.65,64.97,63.96,0.92
+11,data,BinaryMasksDataset,True,True,64,16000,256,512,False,0,0,0.3,0,0.3,0.7,64.59,65.51,64.43,0.76

output/parameter_ensemble_results.csv

@@ -0,0 +1,8 @@
+loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,exp_path,exp_fingerprint,mean_decb_0.1_mean,mean_decb_0.1_std,majority_decb_0.1_mean,majority_decb_0.1_std,mean_decb_0.15_mean,mean_decb_0.15_std,majority_decb_0.15_mean,majority_decb_0.15_std,mean_decb_0.2_mean,mean_decb_0.2_std,majority_decb_0.2_mean,majority_decb_0.2_std,mean_decb_0.25_mean,mean_decb_0.25_std,majority_decb_0.25_mean,majority_decb_0.25_std,mean_decb_0.3_mean,mean_decb_0.3_std,majority_decb_0.3_mean,majority_decb_0.3_std,mean_decb_0.35000000000000003_mean,mean_decb_0.35000000000000003_std,majority_decb_0.35000000000000003_mean,majority_decb_0.35000000000000003_std,mean_decb_0.4_mean,mean_decb_0.4_std,majority_decb_0.4_mean,majority_decb_0.4_std,mean_decb_0.45_mean,mean_decb_0.45_std,majority_decb_0.45_mean,majority_decb_0.45_std,mean_decb_0.5_mean,mean_decb_0.5_std,majority_decb_0.5_mean,majority_decb_0.5_std,mean_decb_0.55_mean,mean_decb_0.55_std,majority_decb_0.55_mean,majority_decb_0.55_std,mean_decb_0.6_mean,mean_decb_0.6_std,majority_decb_0.6_mean,majority_decb_0.6_std,mean_decb_0.65_mean,mean_decb_0.65_std,majority_decb_0.65_mean,majority_decb_0.65_std,mean_decb_0.7000000000000001_mean,mean_decb_0.7000000000000001_std,majority_decb_0.7000000000000001_mean,majority_decb_0.7000000000000001_std,mean_decb_0.75_mean,mean_decb_0.75_std,majority_decb_0.75_mean,majority_decb_0.75_std,mean_decb_0.8_mean,mean_decb_0.8_std,majority_decb_0.8_mean,majority_decb_0.8_std,mean_decb_0.85_mean,mean_decb_0.85_std,majority_decb_0.85_mean,majority_decb_0.85_std,mean_decb_0.9_mean,mean_decb_0.9_std,majority_decb_0.9_mean,majority_decb_0.9_std
+0.0,0.0,0.4,0.0,0.0,0.0,0.0,output/RCC/RCC_eda61b0dbeef45eb9834eb99abf3de47,eda61b0dbeef45eb9834eb99abf3de47,0.5925877364080224,0.006938692176961028,0.6054059130383656,0.005000349574576797,0.6035126188348693,0.004728260141029759,0.6119061094807641,0.004771612617469231,0.6098099529286345,0.005008956024137795,0.6159536245567923,0.004723028783171482,0.6148523779830684,0.004997228208295086,0.6189954489121384,0.0044948481276867736,0.6186538222172048,0.00502869795489423,0.621985204096152,0.004774954625505153,0.6205070300289018,0.005027883757751465,0.623291378755718,0.004327321562583568,0.6225849652375337,0.004950017601162595,0.6236365636313036,0.0036457795039575203,0.6237926423891605,0.004245521464206745,0.6243730233817768,0.003865562101022002,0.6248579763376663,0.0038746170724038623,0.6253689859651296,0.0038256712915678776,0.6255558975895085,0.0035003559632764006,0.6258166240756389,0.003538475265463362,0.6262833942208905,0.0037135461361424717,0.6262065568792188,0.0031894454050451007,0.6273253236863593,0.0036746830982040094,0.6262994361082519,0.0031116644905288416,0.6272132325373873,0.003878365784076129,0.6270689993444903,0.0026541991979802592,0.6272804667385671,0.003956873936047823,0.6267385396602559,0.0033012614217239537,0.6260879001068883,0.004465611244769408,0.6251413300708595,0.004225134164615164,0.6231868306489841,0.004918352621876367,0.6226329677824316,0.004361787090028386,0.6181800184153236,0.005610672783530799,0.6198470439074982,0.004954696486638424
+0.0,0.4,0.0,0.0,0.0,0.0,0.0,output/RCC/RCC_3cb16686fdb05f9648dcaa197d86f0f0,3cb16686fdb05f9648dcaa197d86f0f0,0.5601903136059753,0.023301472469105338,0.60630128982075,0.009031391810457371,0.5912289155240547,0.014539978689266764,0.6139069958869086,0.007412225154103179,0.6061409085401712,0.010197508488928019,0.6199136991759568,0.00617441862334798,0.6156267286921651,0.007407582764683369,0.6243628507068224,0.005386048789697784,0.6219202696488167,0.006023807767133579,0.627993317123969,0.0052411106238195015,0.6266899309276874,0.005782292608949246,0.630758344911048,0.005729210044921515,0.6305298027610053,0.005410864829601481,0.6324871188409406,0.005496168089136652,0.6330042815082161,0.0052029806928775325,0.6341369596175255,0.00570288645199971,0.6348608849106937,0.005737383946614812,0.6367034444955584,0.005948269657512621,0.6356179869960433,0.006651082809640363,0.6379617970567887,0.007621735548929616,0.6373038606955547,0.007482286225428503,0.6373929214121763,0.0075943866595328735,0.6385214379726407,0.007749612447251744,0.6365836881269884,0.007878789646395677,0.6382087597934083,0.008520310354793682,0.6358668030692305,0.008794323282564099,0.6376700544380963,0.009983243894373599,0.633816384081711,0.010045663710479302,0.634780198042358,0.011735836783331111,0.6306598013516426,0.010603940283882294,0.6258786112291219,0.013735296612241604,0.6268541645544079,0.012710301203894917,0.5994669749520999,0.02756344651380324,0.6196971726993523,0.01571137750413604
+0.0,0.0,0.0,0.0,0.4,0.7,1.7,output/RCC/RCC_2af7c45728a8dcfea2f07bb91ce0533b,2af7c45728a8dcfea2f07bb91ce0533b,0.5826360568020215,0.012595617839554981,0.6095929687568431,0.00834218256808553,0.6043880713093237,0.007326381632742741,0.6176269017704928,0.008486659766797773,0.61400139575416,0.00800810836848,0.6232134775311888,0.009164181248383762,0.6204205293978841,0.008675433370333751,0.6268417203203844,0.010032556399304038,0.6251738677814654,0.008872561703807246,0.628747052753139,0.009909835328050468,0.6283415714707669,0.00926429027561411,0.6313819249986645,0.010021579596668924,0.631351169197862,0.009685648338992684,0.6338471793551991,0.009601124092121308,0.6335322333762007,0.009581831429925558,0.6352763894357109,0.009438564262142768,0.6349968861690078,0.009329038783503881,0.636547057474824,0.008930826351155625,0.636303681113632,0.008863004156344424,0.6369896412019351,0.008005211678799438,0.6375303925581131,0.007993016721109783,0.6370158989962398,0.007518244427213655,0.6385391076383669,0.00715861807754437,0.6361470403399192,0.0073215616748080565,0.6385667855095245,0.007054660803734461,0.635115076787719,0.007369483134264981,0.6365988963888245,0.006531500574639648,0.6337739105322636,0.0070301009263275555,0.6333135697356228,0.006484143290302905,0.6308906432555296,0.007420907165245702,0.6264159397395697,0.009725330186215108,0.6276334719050154,0.008654142210449123,0.613726834091754,0.012521610960179504,0.6222409222503507,0.010379731718464326
+0.0,0.0,0.0,0.2,0.0,0.0,0.0,output/RCC/RCC_6b738c9a057c7ccb33fe860a7e794248,6b738c9a057c7ccb33fe860a7e794248,0.5624199226087251,0.02252861628708266,0.6083682901556664,0.005250686725698862,0.596753971224713,0.00878611000090286,0.6164264189857152,0.003787465123051024,0.6114706712495581,0.005052459981180102,0.6217562587608215,0.003283601153607732,0.6200901035473801,0.0039598966713712954,0.6257062255530151,0.0040078324819995,0.6258738500713872,0.0037326679918915054,0.6290224687172388,0.003956192577819221,0.6289799613340611,0.00411760871352514,0.6308312706975496,0.0036189399515947464,0.6318803193438,0.004380888498951218,0.633570204220021,0.0031327108888365694,0.6343843349376913,0.0036621771393077784,0.6351876003287587,0.0028108068330170606,0.6353592153476719,0.0033552770975162007,0.636227058052253,0.002824122049359571,0.6366995883901133,0.003648339910669902,0.6371794747448682,0.0032754494031027666,0.6372310223128566,0.0028274285605381613,0.637397596105933,0.0034806252372786887,0.636989062269214,0.0021750612580496943,0.6367151913788046,0.0033192310294699744,0.6367300011544821,0.0028127376533105333,0.6357072225202722,0.0037938700805259557,0.6357854470448924,0.004282047942908247,0.6345909547100557,0.004725254834295211,0.6319657156794491,0.006100059991395216,0.632275490736644,0.005447795700497811,0.6244529348937902,0.011232212340717949,0.6280154255871647,0.007413008299835295,0.5965931237905145,0.027468497086935904,0.6221765000081014,0.009545952636684774
+0.0,0.0,0.0,0.0,0.0,0.0,0.0,output/RCC/RCC_db8d1771763aa8f126498c42cce309d4,db8d1771763aa8f126498c42cce309d4,0.5,0.0,0.5537879661916825,0.04106878877189942,0.5,0.0,0.5649024105517932,0.041791185990728806,0.5085484519508209,0.019268680907889968,0.5749237895221674,0.041079596505959305,0.5444571477571263,0.03214712864976586,0.5840810042663207,0.0404532896781124,0.5825372218966038,0.03879166318018893,0.5923813374947319,0.03947597050677233,0.6042800251707917,0.03663602709429604,0.5988274074870211,0.038926313889480056,0.6145510341035356,0.03844421835363325,0.6046156266422282,0.038902330760594156,0.623562844631788,0.028778224857791922,0.6105838553408028,0.03973923478188655,0.6358534114907544,0.005153296519285778,0.6157933410181013,0.04047110994332178,0.6343149654521898,0.014769724043567372,0.6035978441222145,0.053715265280667605,0.6268140302314908,0.04274626943355991,0.5890295388391114,0.05815793952578842,0.6260446408630338,0.04255228259256497,0.5796769267830179,0.05400645707842501,0.6161408021518854,0.04061904693905959,0.5641904680381582,0.0507063068719739,0.5882684791471254,0.053732057055781074,0.5498795697762985,0.04484194765543798,0.5331233292480982,0.036693978952032934,0.5358991960466564,0.0404529132129205,0.5,0.0,0.5257042500390874,0.0361009375739564,0.5,0.0,0.5164261313990063,0.030936778498188232
+0.4,0.0,0.0,0.0,0.0,0.0,0.0,output/RCC/RCC_c139bb7b61ed999cb7f2bfbaac9ab7e2,c139bb7b61ed999cb7f2bfbaac9ab7e2,0.5983782205435177,0.009884264679551405,0.6183956210431217,0.008550166839915607,0.6130223716520256,0.007342932744570507,0.6242103282499657,0.006872177501104398,0.6198750178237972,0.007230614285697732,0.6276664043424767,0.006420858225732456,0.6249625798941253,0.006685542446485032,0.6307023143737753,0.0057848050660980815,0.6289401634687966,0.0065633778887345924,0.6328332848947091,0.0057432831539235545,0.6315292614965042,0.005822015192498231,0.635060529295631,0.0054843609229875325,0.6333671811611684,0.005474931339230299,0.6364836154709319,0.005172056322599254,0.6351046917121231,0.005041857657333599,0.6371665991307813,0.004959844238128174,0.6367498070443556,0.004929208808637456,0.6375454175539093,0.004909028040595848,0.6379770100177541,0.004947809892656639,0.6382716566983099,0.005170085348550484,0.6387987833992019,0.004468683098999042,0.6381772822063562,0.004848158648201774,0.6389046970728899,0.004518472333979729,0.6380386259400279,0.004703188350964925,0.6392087391924257,0.004828924529631297,0.6380997146200023,0.0046603068067981104,0.6390258998334366,0.004616112524019607,0.6377271967893444,0.004623803691325905,0.6378830246137005,0.003989505785701553,0.6363657989029075,0.004277037516932451,0.6359403021494983,0.003887384997503192,0.6344168538517805,0.004360736148762412,0.6283114735480565,0.006784927171628257,0.631382177811817,0.006242860112616115
+0.4,0.4,0.4,0.2,0.4,0.7,1.7,output/RCC/RCC_1153122048000b25de26fda369342ae0,1153122048000b25de26fda369342ae0,0.5132975686367033,0.016221988890468775,0.5814105323575836,0.016436566069371245,0.5503070364046805,0.02799579758367218,0.594308479526063,0.01395587863664963,0.5785190867163398,0.023253188960347296,0.6033877065817095,0.011756227977684285,0.5978159781897248,0.014112897743266326,0.6111292898491703,0.008758621576557284,0.6102343951974448,0.008414594723601775,0.616592223061424,0.0077033632311362095,0.6178160342033452,0.004731202904364016,0.6210472889163368,0.0065499035883224265,0.6237185784735597,0.0030994549398435742,0.624101159959374,0.005329038944462259,0.6271758782193217,0.004020251905364154,0.6272760223021425,0.005627471396119683,0.6304623242051057,0.004699604386425982,0.63014738668454,0.006310525897322494,0.632735057116788,0.005293424764128294,0.6323061563141653,0.00506133140573713,0.6334494046447622,0.0050696322887465416,0.6309316516164836,0.0064820507798342385,0.6324917907152199,0.0055211399319096055,0.6282236214474725,0.007604028164980533,0.6289383702810827,0.0073212656870781665,0.6243637858335389,0.010835436806515065,0.6226400991809493,0.011722212458924457,0.6190908628514775,0.014623854123354815,0.6099207038859504,0.020614209671051823,0.6144870676768216,0.01832210682604785,0.5824997247250099,0.03322203919549863,0.608232693022466,0.02210615709870419,0.5397607872552699,0.028741371303585152,0.5968847837045327,0.026674431554009028

output/results.csv

@@ -0,0 +1,8 @@
+loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,exp_path,exp_fingerprint,mean_decb_0.1_mean,mean_decb_0.1_std,majority_decb_0.1_mean,majority_decb_0.1_std,mean_decb_0.15_mean,mean_decb_0.15_std,majority_decb_0.15_mean,majority_decb_0.15_std,mean_decb_0.2_mean,mean_decb_0.2_std,majority_decb_0.2_mean,majority_decb_0.2_std,mean_decb_0.25_mean,mean_decb_0.25_std,majority_decb_0.25_mean,majority_decb_0.25_std,mean_decb_0.3_mean,mean_decb_0.3_std,majority_decb_0.3_mean,majority_decb_0.3_std,mean_decb_0.35000000000000003_mean,mean_decb_0.35000000000000003_std,majority_decb_0.35000000000000003_mean,majority_decb_0.35000000000000003_std,mean_decb_0.4_mean,mean_decb_0.4_std,majority_decb_0.4_mean,majority_decb_0.4_std,mean_decb_0.45_mean,mean_decb_0.45_std,majority_decb_0.45_mean,majority_decb_0.45_std,mean_decb_0.5_mean,mean_decb_0.5_std,majority_decb_0.5_mean,majority_decb_0.5_std,mean_decb_0.55_mean,mean_decb_0.55_std,majority_decb_0.55_mean,majority_decb_0.55_std,mean_decb_0.6_mean,mean_decb_0.6_std,majority_decb_0.6_mean,majority_decb_0.6_std,mean_decb_0.65_mean,mean_decb_0.65_std,majority_decb_0.65_mean,majority_decb_0.65_std,mean_decb_0.7000000000000001_mean,mean_decb_0.7000000000000001_std,majority_decb_0.7000000000000001_mean,majority_decb_0.7000000000000001_std,mean_decb_0.75_mean,mean_decb_0.75_std,majority_decb_0.75_mean,majority_decb_0.75_std,mean_decb_0.8_mean,mean_decb_0.8_std,majority_decb_0.8_mean,majority_decb_0.8_std,mean_decb_0.85_mean,mean_decb_0.85_std,majority_decb_0.85_mean,majority_decb_0.85_std,mean_decb_0.9_mean,mean_decb_0.9_std,majority_decb_0.9_mean,majority_decb_0.9_std
+0.0,0.0,0.0,0.0,0.0,0.0,0.0,output/BCC/BCC_db8d1771763aa8f126498c42cce309d4,db8d1771763aa8f126498c42cce309d4,0.49998747024182427,2.5059516351344513e-05,0.5311949955175945,0.015112175122027827,0.5071166723864466,0.003570711297014624,0.5492620919778984,0.01851216603386002,0.5293378714515982,0.009879962150055866,0.5638004619281626,0.01942253972659972,0.553531742411179,0.014899927077776539,0.5757971809585419,0.018554233762557208,0.5754232082774748,0.0158392615728627,0.5874900683903749,0.016891075996529916,0.5949602384120131,0.01533473801319841,0.5982690917256116,0.015698837581181586,0.6115475837772977,0.013087107406046813,0.6088634872132747,0.013455978264132511,0.6254632852710154,0.010653191920014791,0.6192807856778662,0.01089874115168,0.6372496506009454,0.008131179811915431,0.6313621567018248,0.010903739879631091,0.646534199529463,0.006163372309844989,0.6430586211108267,0.008147159053007483,0.6524925252831009,0.005238623995814059,0.6476689678114534,0.00765243991104426,0.6545960773248105,0.006199612336030588,0.647929208407622,0.008051866737801753,0.6536517638106375,0.007702468341359227,0.6460285619465342,0.008419549087757138,0.6465563615627319,0.009655918108778926,0.639586622488837,0.009224189389706849,0.6279623883952109,0.012026165828036067,0.6243279671909054,0.011623022983736532,0.584630905766933,0.0165077441526688,0.6003664756902956,0.013317932476419391,0.5150911783322174,0.011563903991300705,0.5643390873982775,0.015885951694295016
+0.0,0.0,0.4,0.0,0.0,0.0,0.0,output/BCC/BCC_eda61b0dbeef45eb9834eb99abf3de47,eda61b0dbeef45eb9834eb99abf3de47,0.5822832349016133,0.014277229463365937,0.614223574343401,0.007123043977676212,0.6019933386647773,0.01087559261092599,0.6207661108799766,0.006582369701223912,0.6136665724585239,0.008045625076395373,0.6255756019781905,0.005612628017794035,0.6210624755621785,0.006706298557609762,0.6295592493092187,0.0049612343822471785,0.6264034381761694,0.005703521664488483,0.6326191155013259,0.004889324894579935,0.6307692262146901,0.005080289638922534,0.6357464499112485,0.004557767212874401,0.6346350682556039,0.004524266148547106,0.638249092419579,0.004679185634818143,0.637112206709958,0.0039821643350878705,0.6399670085414757,0.004817516942879441,0.6398018379083867,0.004480682403048495,0.6414604538388935,0.005114714037223655,0.6419113540849135,0.004404764498926187,0.6430059767660765,0.005530397909719675,0.643286044868113,0.004808962693447943,0.6441961891340118,0.00528612884781965,0.6447250864667973,0.005041608762556574,0.6446017991725247,0.005366178133230501,0.64539916715641,0.004539668430481869,0.6447992142905556,0.004774407361312133,0.645656430384299,0.004498939376424698,0.6450833884917122,0.004451675740868208,0.6443659949881717,0.00374981901241002,0.643589012766707,0.004048860065531186,0.6406433072200841,0.004001505707370434,0.6416505364356759,0.0038211772856512295,0.6316921145190574,0.0066680056949816775,0.6369806414296361,0.0038874311762615438
+0.4,0.0,0.0,0.0,0.0,0.0,0.0,output/BCC/BCC_c139bb7b61ed999cb7f2bfbaac9ab7e2,c139bb7b61ed999cb7f2bfbaac9ab7e2,0.5953526034848817,0.009994961530947009,0.6213078650303321,0.00591329204483482,0.6105428037478601,0.00826912741156256,0.6291588675218857,0.005729586082042761,0.6209517246048625,0.0069540219030815925,0.6340199785546778,0.005783324183232921,0.6282333928167908,0.006316000816135033,0.6384810010258011,0.0055530736750792715,0.6341372321670208,0.005565533421336869,0.6422332923682044,0.005370511654233748,0.6387102442859247,0.0057379267615629,0.6451112759440335,0.004926673673350581,0.6428876367150561,0.005291873545943229,0.6471772584475588,0.004731957811971279,0.6463076244349156,0.004599532710726014,0.6492062503456208,0.004344008756507752,0.6494991771448789,0.004511703754646507,0.6508922518614473,0.0043329070030276445,0.6519152532343824,0.004496989231629901,0.6525704960547427,0.004161995712489743,0.6534563758969557,0.003935674790074506,0.6539281196005831,0.004000697056933606,0.6553677166363423,0.004262669730173998,0.6549192242241391,0.003914594026828055,0.6572893635059535,0.00409932082129434,0.6558723504852677,0.0038505469952177023,0.6586411075013027,0.00377301934528889,0.6567613655865022,0.003719584656560209,0.6584077816377052,0.003415705707993253,0.6569675859328822,0.004009033739026109,0.6559369277987817,0.004188169286718383,0.6565178971223773,0.004311121199108305,0.6484983547220778,0.006610219271521315,0.654074583000212,0.004369127397296004
+0.0,0.4,0.0,0.0,0.0,0.0,0.0,output/BCC/BCC_3cb16686fdb05f9648dcaa197d86f0f0,3cb16686fdb05f9648dcaa197d86f0f0,0.5690024628324254,0.010927333003299877,0.6287383189515544,0.011210983071878919,0.604468122855388,0.009148452718499626,0.6441459554425559,0.008713713824424095,0.627987303178382,0.008082376594374073,0.6546106530838525,0.007586574662152933,0.6447440098602017,0.006608820210232434,0.6621088082996998,0.005996089788957881,0.6567322920739638,0.006363534995890463,0.6675398654387429,0.004987676358600798,0.665151577180021,0.005230404192603345,0.6719235458817151,0.004324195591334658,0.6712674478198353,0.0034716996960555213,0.6745632977030318,0.004088170605283382,0.6762623797149794,0.003367696424914243,0.6757873193659425,0.005054862755228786,0.6793676742689932,0.003673956074037863,0.676116442617824,0.006037872342597853,0.679912942428186,0.004407342862349408,0.6755556109426253,0.00614135931317807,0.6796134697419122,0.005705103065690281,0.673469599810501,0.007519687092551546,0.6773784915141057,0.007545441632784105,0.6701527260231349,0.00850397549736509,0.6732220214879525,0.009277755800843165,0.6652025063426026,0.009688479374158034,0.6651722477096726,0.011229497924456265,0.6584605613173771,0.01011780551000107,0.6544883084349031,0.011918426401518013,0.6470116865088338,0.010894085280778626,0.6357881208494237,0.012967137635907363,0.6314928855371441,0.01122065967116801,0.606838112704893,0.011140499621092707,0.6099331847731823,0.011661629429919473
+0.0,0.0,0.0,0.0,0.4,0.7,1.7,output/BCC/BCC_2af7c45728a8dcfea2f07bb91ce0533b,2af7c45728a8dcfea2f07bb91ce0533b,0.5936077293984254,0.00868807871395221,0.6269160567854806,0.011130936303142557,0.6158427890151501,0.007983391241809284,0.635995843977331,0.009098095730947305,0.6292450439129869,0.007659533059946964,0.6431906976564089,0.006793202363809304,0.6383004943145612,0.00621909168042349,0.6480246965932027,0.0057221895464368636,0.6448467384640982,0.00557737758267446,0.6515003370403447,0.004127659321033492,0.6501624404390068,0.004591849315864673,0.6546795310404364,0.003719573480123698,0.6537520291309924,0.0039007373669972803,0.6572714513651591,0.0034780252029153733,0.6577350693345239,0.0032173447139593934,0.6593395935884779,0.003312215922221381,0.6602243484591537,0.003125371825693809,0.6617596338633989,0.003120773801126948,0.6618745228739138,0.0030169744092991098,0.6629351868277918,0.0032040814077269734,0.6639818154224895,0.002777445398186179,0.6637310069185093,0.003025302773704986,0.6651338614628284,0.0026472363538318625,0.6639097984475806,0.003201896586323688,0.6657409400882451,0.002789570677155023,0.6635400474642126,0.003334167029839661,0.6647573296359833,0.003227312693467614,0.6629372046338828,0.003994880893117735,0.6617911866365186,0.0047128503364520504,0.6606628279667962,0.00488589728178858,0.6547629903950175,0.007557916066729358,0.6572753770177169,0.0067842262936338256,0.6383329089079812,0.012128510426312387,0.650401946597366,0.009125878443279034
+0.4,0.4,0.4,0.2,0.4,0.7,1.7,output/BCC/BCC_1153122048000b25de26fda369342ae0,1153122048000b25de26fda369342ae0,0.5225303077858218,0.007568159447105149,0.5865186014278662,0.0184120161380992,0.562724566216802,0.012901497073023667,0.6073839701348909,0.01648127704025501,0.5941562127018365,0.014926857289822336,0.6218910034719799,0.013627611288937172,0.6148360428700432,0.012863575358821473,0.6316475047454626,0.011613506431628635,0.6304329687448134,0.010688307414604904,0.6395360711362453,0.009944167710997405,0.6417317937783003,0.008282618266259057,0.6464192719860885,0.008427952061924406,0.6507958614430545,0.0062516193420844475,0.650998707814648,0.005352543253607115,0.6564640864537524,0.00444612155082319,0.6537926145703156,0.003353268510819472,0.6578958631986491,0.003898418377417333,0.6538147249131628,0.0031009441978869035,0.6583235102735934,0.004066073278074313,0.6526691815278508,0.004476863187575506,0.6562920062210456,0.00647013527289821,0.6510962579781345,0.006263074435767392,0.6526498295738608,0.008036407827669992,0.6471688817797956,0.007860935743611026,0.6453475951529498,0.009550717987239828,0.6401940761047663,0.011552699537804668,0.6353232303558636,0.010542149569327735,0.630032102007344,0.013185635642050351,0.6201749857982917,0.011957620145608005,0.614819189912977,0.01313708163536355,0.5979462051580425,0.011662348183286715,0.5977080495294236,0.01193889674066375,0.563311664144738,0.008927272448876294,0.5765971409821097,0.010734152991207994
+0.0,0.0,0.0,0.2,0.0,0.0,0.0,output/BCC/BCC_6b738c9a057c7ccb33fe860a7e794248,6b738c9a057c7ccb33fe860a7e794248,0.5766728918851037,0.012111414191176301,0.619254110600886,0.007654705960053367,0.6020821418691173,0.011080340865239754,0.6267523513408852,0.006327976103732864,0.6163851455938728,0.008784340490742249,0.6318152044499025,0.005101135945543378,0.6258596032513913,0.00671138459718275,0.6361671582519737,0.004128458886835169,0.6326851542440298,0.005253900331735938,0.6389294191917625,0.004134210503546723,0.6377766335084831,0.003699374340120643,0.6422689897733038,0.003959943184768117,0.6414004844915014,0.0028319819209817825,0.6448580577265846,0.0037535263170551205,0.6443489728624534,0.0030926918528754546,0.646690443696669,0.0037049438148316336,0.6460562153822198,0.0029605032692547194,0.6475817378755793,0.0038049288869173433,0.6478194141930168,0.00367707954524727,0.6483851526989561,0.0037019209751211857,0.6489051122879486,0.0036247132467559185,0.6495948824302301,0.003865228231370226,0.649887835544612,0.00413608853741559,0.6504804549493917,0.0039531019557641316,0.6512579935477572,0.0037483827055999333,0.6501346131355399,0.003427444327419835,0.650459772202004,0.002921437670037661,0.6492172134141118,0.0023493161281014445,0.6484262409451069,0.0024546228019007646,0.6472426101775696,0.002936534871812297,0.6421825060591513,0.0043530779605389985,0.6438633357133483,0.0036908043943732283,0.6270617974590342,0.009282636418283205,0.637010962091072,0.006823800072291407

output/secondrun/bcc_metrics.csv

@@ -0,0 +1,71 @@
+worker,root,class_name,use_preprocessed,n_mels,sr,hop_length,n_fft,stretch,loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,max,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.1,65.816,0.38641371,None
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,66.19,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.78,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.9,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,66.17,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,66.13,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.18,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.77,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,66.02,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.92,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,65.11,64.479,0.461168805,Noise
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.47,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.79,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.56,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.44,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.46,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.29,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.48,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.09,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,65.1,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.95,65.646,0.499893322,Loudness
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,66,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.14,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.84,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,66.27,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.27,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.05,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.43,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,66.38,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.13,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.85,68.186,0.444727132,Shift
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,67.6,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.03,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.22,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.66,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.56,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.28,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,67.89,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,68.29,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,67.48,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.07,66.394,0.342967443,Speed
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.46,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.89,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,67,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.31,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.56,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.31,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.71,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.02,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,66.61,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.17,66.348,0.385625034,All
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.63,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.05,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.45,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.03,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.71,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,66.52,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.86,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,67.06,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.14,65.083,0.323077218,Mask
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.31,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.18,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.37,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.04,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.18,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.18,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.4,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.4,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.63,,,

output/secondrun/bcmc_metrics.csv

@@ -0,0 +1,71 @@
+worker,root,class_name,use_preprocessed,n_mels,sr,hop_length,n_fft,stretch,loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,max,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,60.29,61.349,0.670496499,noise
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,61.95,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,62.12,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,61.33,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,61.73,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,60.74,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,61.09,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,60.82,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,62.36,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,61.06,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.21,63.542,0.90676225,speed
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.9,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.72,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.88,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.2,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,62.72,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.12,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,62.77,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.45,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.45,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.23,63.488,0.788117307,all
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.7,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,62.52,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.4,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,62.51,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.39,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,62.34,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.7,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.66,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.43,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.61,62.268,0.334524538,mask
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.57,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.74,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,61.75,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.22,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.12,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.49,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.15,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,61.8,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,62.23,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.35,63.541,0.255884088,none
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.28,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.78,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.47,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.16,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.57,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.86,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.9,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.67,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.37,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.65,66.309,0.693468737,shift
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.8,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.64,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.75,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.74,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.78,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.41,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.6,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,64.81,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.91,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,61.44,62.869,0.640979286,loudness
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.03,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.33,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,62.98,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,62.46,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,62.56,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.13,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.72,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.41,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,62.63,,,

output/secondrun/cc_metrics.csv

@@ -0,0 +1,71 @@
+worker,root,class_name,use_preprocessed,n_mels,sr,hop_length,n_fft,stretch,loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,max,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.84,64.714,0.302294926,None
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.67,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.45,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.88,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.19,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.08,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.87,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.58,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.44,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.14,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.6,64.784,0.403517864,Speed
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.95,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.86,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.69,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.79,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.1,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.91,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.4,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.5,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.04,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.17,65.029,0.406760918,All
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.34,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.33,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.27,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.21,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.29,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.37,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,65.09,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.98,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.24,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.77,63.618,0.224192378,noise
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.65,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.57,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.37,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.98,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.19,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.61,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.77,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.52,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.75,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.77,64.305,0.435488232,loudness
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,65.03,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.83,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.97,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.72,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.43,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.11,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.45,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.84,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.9,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.18,64.268,0.335949732,mask
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.3,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.51,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.13,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,63.71,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.01,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.24,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.38,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.22,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.1,65.394,0.54946843,shift
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,66.07,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.32,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.86,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.21,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.24,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,64.23,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.54,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.28,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.09,,,

output/secondrun/rcc_metrics.csv

@@ -0,0 +1,71 @@
+worker,root,class_name,use_preprocessed,n_mels,sr,hop_length,n_fft,stretch,loudness_ratio,shift_ratio,noise_ratio,mask_ratio,speed_amount,speed_min,speed_max,max,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.3,63.38,0.406639343,noise
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.27,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.54,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,62.69,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,64.05,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.71,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,62.8,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.55,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.58,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0.4,0,0,0,0,63.31,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,63.6,64.377,0.775901626,shift
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.57,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.39,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,64.55,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,63.87,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,64.15,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,63.76,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,64.17,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,63.46,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0.4,0,0,0,0,0,65.25,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.62,64.069,0.766571733,speed
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.32,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,62.61,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,65.44,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.48,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.26,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.05,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.66,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,63.99,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0,0.4,0.7,1.7,64.26,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,63.75,64.304,0.449102809,mask
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.02,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.31,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.54,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.23,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.52,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.25,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,63.79,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,65.33,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0,0,0,0.2,0,0,0,64.3,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.64,64.53,0.53264539,none
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.19,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.2,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.4,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.88,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.67,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,63.85,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.25,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,65.5,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,FALSE,0,0,0,0,0,0,0,64.72,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.84,64.14,0.372618363,loudness
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.5,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.38,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.37,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.9,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.11,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.24,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,63.97,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0,0,0,0,0,0,64.09,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.32,64.118,0.417047826,all
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.58,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.55,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.54,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.04,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.64,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,63.48,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.41,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.24,,,
+11,data,BinaryMasksDataset,FALSE,64,16000,256,512,TRUE,0.4,0.4,0.4,0.2,0.4,0.7,1.7,64.38,,,

repair_outputs.py

@@ -0,0 +1,45 @@
+import csv
+from pathlib import Path
+
+import shutil
+
+if __name__ == '__main__':
+
+    for old_out_file in (Path() / 'output').rglob('*_test_out.csv'):
+        old_out_file.unlink()
+
+    for new_out_file in (Path() / 'output').rglob('*_test_out_repair.csv'):
+        shutil.move(str(new_out_file), str(new_out_file).replace('_test_out_repair', '_test_out'))
+    exit()
+    '''
+    with old_out_file.open('r') as old_f:
+        predictions = []
+        file_names = []
+
+        idx = 1
+        zeros = '00000'
+
+        _ = old_f.readline()
+
+        for row in old_f:
+            split_row = row.split(',')
+            file_names.append(f'test_{zeros[:-len(str(idx))]}{idx}.wav')
+            predictions.append(split_row[-1].strip()
+                               .replace('"', '').replace('(', '').replace(')', '').replace("'", '')
+                               )
+            idx += 1
+    try:
+        (old_out_file.parent / f'{old_out_file.name}_repair').unlink()
+    except FileNotFoundError:
+        pass
+    with (old_out_file.parent / f'{old_out_file.name[:-4]}_repair.csv').open('w') as new_f:
+        headers = ['file_name', 'prediction']
+
+        writer = csv.DictWriter(new_f, delimiter=',', lineterminator='\n', fieldnames=headers)
+        writer.writeheader()  # write a header
+
+        writer.writerows([dict(file_name=file_name, prediction=prediction)
+                          for file_name, prediction in zip(file_names, predictions)]
+                         )
+    '''
+    pass

util/module_mixins.py

@@ -1,11 +1,17 @@
-from collections import defaultdict
-
+# Imports from python Internals
 from abc import ABC
-from argparse import Namespace
+from itertools import cycle
+from collections import defaultdict, namedtuple
 
-import sklearn
-import torch
+# Numerical Imports, Metrics and Plotting
 import numpy as np
+from sklearn.ensemble import IsolationForest
+from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay, roc_auc_score, roc_curve, auc, f1_score, \
+    recall_score, average_precision_score
+from matplotlib import pyplot as plt
+
+# Import Deep Learning Framework
+import torch
 from torch import nn
 from torch.optim import Adam
 from torch.optim.lr_scheduler import CosineAnnealingWarmRestarts
@@ -13,15 +19,25 @@ from torch.utils.data import DataLoader
 from torchcontrib.optim import SWA
 from torchvision.transforms import Compose, RandomApply
 
-from ml_lib.audio_toolset.audio_augmentation import Speed
+# Import Functions and Modules from MLLIB
 from ml_lib.audio_toolset.mel_augmentation import NoiseInjection, LoudnessManipulator, ShiftTime, MaskAug
-from ml_lib.audio_toolset.audio_io import AudioToMel, MelToImage, NormalizeLocal
+from ml_lib.audio_toolset.audio_io import NormalizeLocal
 from ml_lib.modules.util import LightningBaseModule
+from ml_lib.utils.tools import to_one_hot
 from ml_lib.utils.transforms import ToTensor
 
+# Import Project Variables
 import variables as V
 
 
+class BaseLossMixin:
+
+    absolute_loss = nn.L1Loss()
+    nll_loss = nn.NLLLoss()
+    bce_loss = nn.BCELoss()
+    ce_loss = nn.CrossEntropyLoss()
+
+
 class BaseOptimizerMixin:
 
     def configure_optimizers(self):
@@ -60,16 +76,12 @@ class BaseOptimizerMixin:
 
 class BaseTrainMixin:
 
-    absolute_loss = nn.L1Loss()
-    nll_loss = nn.NLLLoss()
-    bce_loss = nn.BCELoss()
-
     def training_step(self, batch_xy, batch_nb, *args, **kwargs):
         assert isinstance(self, LightningBaseModule)
         batch_x, batch_y = batch_xy
         y = self(batch_x).main_out
-        bce_loss = self.bce_loss(y.squeeze(), batch_y)
-        return dict(loss=bce_loss)
+        loss = self.ce_loss(y.squeeze(), batch_y.long())
+        return dict(loss=loss)
 
     def training_epoch_end(self, outputs):
         assert isinstance(self, LightningBaseModule)
@@ -84,55 +96,39 @@ class BaseTrainMixin:
 
 class BaseValMixin:
 
-    absolute_loss = nn.L1Loss()
-    nll_loss = nn.NLLLoss()
-    bce_loss = nn.BCELoss()
-
-    def validation_step(self, batch_xy, batch_idx, dataloader_idx, *args, **kwargs):
+    def validation_step(self, batch_xy, batch_idx, *args, **kwargs):
         assert isinstance(self, LightningBaseModule)
         batch_x, batch_y = batch_xy
         y = self(batch_x).main_out
-        val_bce_loss = self.bce_loss(y.squeeze(), batch_y)
-        return dict(val_bce_loss=val_bce_loss,
+        val_loss = self.ce_loss(y.squeeze(), batch_y.long())
+        return dict(val_loss=val_loss,
                     batch_idx=batch_idx, y=y, batch_y=batch_y)
 
     def validation_epoch_end(self, outputs, *_, **__):
         assert isinstance(self, LightningBaseModule)
         summary_dict = dict()
-        for output_idx, output in enumerate(outputs):
-            keys = list(output[0].keys())
-            ident = '' if output_idx == 0 else '_train'
-            summary_dict.update({f'mean{ident}_{key}': torch.mean(torch.stack([output[key]
-                                                                               for output in output]))
-                                 for key in keys if 'loss' in key}
-                                )
 
-            # UnweightedAverageRecall
-            y_true = torch.cat([output['batch_y'] for output in output]) .cpu().numpy()
-            y_pred = torch.cat([output['y'] for output in output]).squeeze().cpu().numpy()
+        keys = list(outputs[0].keys())
+        summary_dict.update({f'mean_{key}': torch.mean(torch.stack([output[key]
+                                                                    for output in outputs]))
+                             for key in keys if 'loss' in key}
+                            )
 
-            y_pred = (y_pred >= 0.5).astype(np.float32)
+        additional_scores = self.additional_scores(outputs)
+        summary_dict.update(**additional_scores)
 
-            uar_score = sklearn.metrics.recall_score(y_true, y_pred, labels=[0, 1], average='macro',
-                                                     sample_weight=None, zero_division='warn')
-            uar_score = torch.as_tensor(uar_score)
-            summary_dict.update({f'uar{ident}_score': uar_score})
-            for key in summary_dict.keys():
-                self.log(key, summary_dict[key])
+        for key in summary_dict.keys():
+            self.log(key, summary_dict[key])
 
 
 class BaseTestMixin:
 
-    absolute_loss = nn.L1Loss()
-    nll_loss = nn.NLLLoss()
-    bce_loss = nn.BCELoss()
-
-    def test_step(self, batch_xy, batch_idx, dataloader_idx, *args, **kwargs):
+    def test_step(self, batch_xy, batch_idx, *_, **__):
         assert isinstance(self, LightningBaseModule)
         batch_x, batch_y = batch_xy
         y = self(batch_x).main_out
-        test_bce_loss = self.bce_loss(y.squeeze(), batch_y)
-        return dict(test_bce_loss=test_bce_loss,
+        test_loss = self.ce_loss(y.squeeze(), batch_y.long())
+        return dict(test_loss=test_loss,
                     batch_idx=batch_idx, y=y, batch_y=batch_y)
 
     def test_epoch_end(self, outputs, *_, **__):
@@ -145,16 +141,9 @@ class BaseTestMixin:
                              for key in keys if 'loss' in key}
                             )
 
-        # UnweightedAverageRecall
-        y_true = torch.cat([output['batch_y'] for output in outputs]) .cpu().numpy()
-        y_pred = torch.cat([output['y'] for output in outputs]).squeeze().cpu().numpy()
+        additional_scores = self.additional_scores(outputs)
+        summary_dict.update(**additional_scores)
 
-        y_pred = (y_pred >= 0.5).astype(np.float32)
-
-        uar_score = sklearn.metrics.recall_score(y_true, y_pred, labels=[0, 1], average='macro',
-                                                 sample_weight=None, zero_division='warn')
-        uar_score = torch.as_tensor(uar_score)
-        summary_dict.update({f'uar_score': uar_score})
         for key in summary_dict.keys():
             self.log(key, summary_dict[key])
 
@@ -167,53 +156,56 @@ class DatasetMixin:
         # Dataset
         # =============================================================================
         # Mel Transforms
-        mel_transforms = Compose([
-            # Audio to Mel Transformations
-            AudioToMel(sr=self.params.sr,
-                       n_mels=self.params.n_mels,
-                       n_fft=self.params.n_fft,
-                       hop_length=self.params.hop_length),
-            MelToImage()])
-
-        mel_transforms_train = Compose([
-            # Audio to Mel Transformations
-            Speed(max_amount=self.params.speed_amount,
-                  speed_min=self.params.speed_min,
-                  speed_max=self.params.speed_max
-                  ),
-            mel_transforms])
+        mel_kwargs = dict(sample_rate=self.params.sr,
+                          n_mels=self.params.n_mels,
+                          n_fft=self.params.n_fft,
+                          hop_length=self.params.hop_length)
 
         # Utility
-        util_transforms = Compose([NormalizeLocal(), ToTensor()])
+        utility_transforms = Compose([NormalizeLocal(), ToTensor()])
 
         # Data Augmentations
-        aug_transforms = Compose([
+        mel_augmentations = Compose([
             RandomApply([
-                NoiseInjection(self.params.noise_ratio),
-                LoudnessManipulator(self.params.loudness_ratio),
-                ShiftTime(self.params.shift_ratio),
-                MaskAug(self.params.mask_ratio),
+                NoiseInjection(0.2),
+                LoudnessManipulator(0.5),
+                ShiftTime(0.4),
+                MaskAug(0.2),
             ], p=0.6),
-            util_transforms])
+            utility_transforms])
 
         # Datasets
-        dataset = Namespace(
-            **dict(
-                # TRAIN DATASET
-                train_dataset=self.dataset_class(self.params.root, setting=V.DATA_OPTIONS.train,
-                                                 use_preprocessed=self.params.use_preprocessed,
-                                                 stretch_dataset=self.params.stretch,
-                                                 mel_transforms=mel_transforms_train, transforms=aug_transforms),
-                # VALIDATION DATASET
-                val_train_dataset=self.dataset_class(self.params.root, setting=V.DATA_OPTIONS.train,
-                                                     mel_transforms=mel_transforms, transforms=util_transforms),
-                val_dataset=self.dataset_class(self.params.root, setting=V.DATA_OPTIONS.devel,
-                                               mel_transforms=mel_transforms, transforms=util_transforms),
-                # TEST DATASET
-                test_dataset=self.dataset_class(self.params.root, setting=V.DATA_OPTIONS.test,
-                                                mel_transforms=mel_transforms, transforms=util_transforms),
-            )
-        )
+        Dataset = namedtuple('Datasets', 'train_dataset val_dataset test_dataset')
+        dataset = Dataset(self.dataset_class(data_root=self.params.root,                # TRAIN DATASET
+                                             setting=V.DATA_OPTION_train,
+                                             fold=list(range(1,8)),
+                                             reset=self.params.reset,
+                                             mel_kwargs=mel_kwargs,
+                                             mel_augmentations=mel_augmentations),
+                          val_dataset=self.dataset_class(data_root=self.params.root,    # VALIDATION DATASET
+                                                         setting=V.DATA_OPTION_devel,
+                                                         fold=9,
+                                                         reset=self.params.reset,
+                                                         mel_kwargs=mel_kwargs,
+                                                         mel_augmentations=utility_transforms),
+                          test_dataset=self.dataset_class(data_root=self.params.root,    # TEST DATASET
+                                                          setting=V.DATA_OPTION_test,
+                                                          fold=10,
+                                                          reset=self.params.reset,
+                                                          mel_kwargs=mel_kwargs,
+                                                          mel_augmentations=utility_transforms),
+                          )
+
+        if dataset.train_dataset.task_type == V.TASK_OPTION_binary:
+        # noinspection PyAttributeOutsideInit
+            self.additional_scores = BinaryScores(self)
+
+        elif dataset.train_dataset.task_type == V.TASK_OPTION_multiclass:
+        # noinspection PyAttributeOutsideInit
+            self.additional_scores = MultiClassScores(self)
+        else:
+            raise ValueError
+
         return dataset
 
 
@@ -240,10 +232,185 @@ class BaseDataloadersMixin(ABC):
     # Validation Dataloader
     def val_dataloader(self):
         assert isinstance(self, LightningBaseModule)
-        val_dataloader = DataLoader(dataset=self.dataset.val_dataset, shuffle=False, pin_memory=True,
-                                    batch_size=self.params.batch_size, num_workers=self.params.worker)
+        return DataLoader(dataset=self.dataset.val_dataset, shuffle=False, pin_memory=True,
+                          batch_size=self.params.batch_size, num_workers=self.params.worker)
 
-        train_dataloader = DataLoader(self.dataset.val_train_dataset, num_workers=self.params.worker,
-                                      pin_memory=True,
-                                      batch_size=self.params.batch_size, shuffle=False)
-        return [val_dataloader, train_dataloader]
+
+class BaseScores(ABC):
+
+    def __init__(self, lightning_model):
+        self.model = lightning_model
+        pass
+
+    def __call__(self, outputs):
+        # summary_dict = dict()
+        # return summary_dict
+        raise NotImplementedError
+
+
+class MultiClassScores(BaseScores):
+
+    def __init__(self, *args):
+        super(MultiClassScores, self).__init__(*args)
+        pass
+
+    def __call__(self, outputs):
+        summary_dict = dict()
+        #######################################################################################
+        # Additional Score  -  UAR  -  ROC  -  Conf. Matrix  -  F1
+        #######################################################################################
+        #
+        # INIT
+        y_true = torch.cat([output['batch_y'] for output in outputs]).cpu().numpy()
+        y_true_one_hot = to_one_hot(y_true, self.model.n_classes)
+
+        y_pred = torch.cat([output['y'] for output in outputs]).squeeze().cpu().float().numpy()
+        y_pred_max = np.argmax(y_pred, axis=1)
+
+        class_names = {val: key for key, val in self.model.dataset.test_dataset.classes.items()}
+        ######################################################################################
+        #
+        # F1 SCORE
+        micro_f1_score = f1_score(y_true, y_pred_max, labels=None, pos_label=1, average='micro', sample_weight=None,
+                                  zero_division=True)
+        macro_f1_score = f1_score(y_true, y_pred_max, labels=None, pos_label=1, average='macro', sample_weight=None,
+                                  zero_division=True)
+        summary_dict.update(dict(micro_f1_score=micro_f1_score, macro_f1_score=macro_f1_score))
+
+        #######################################################################################
+        #
+        # ROC Curve
+
+        # Compute ROC curve and ROC area for each class
+        fpr = dict()
+        tpr = dict()
+        roc_auc = dict()
+        for i in range(self.model.n_classes):
+            fpr[i], tpr[i], _ = roc_curve(y_true_one_hot[:, i], y_pred[:, i])
+            roc_auc[i] = auc(fpr[i], tpr[i])
+
+        # Compute micro-average ROC curve and ROC area
+        fpr["micro"], tpr["micro"], _ = roc_curve(y_true_one_hot.ravel(), y_pred.ravel())
+        roc_auc["micro"] = auc(fpr["micro"], tpr["micro"])
+
+        # First aggregate all false positive rates
+        all_fpr = np.unique(np.concatenate([fpr[i] for i in range(self.model.n_classes)]))
+
+        # Then interpolate all ROC curves at this points
+        mean_tpr = np.zeros_like(all_fpr)
+        for i in range(self.model.n_classes):
+            mean_tpr += np.interp(all_fpr, fpr[i], tpr[i])
+
+        # Finally average it and compute AUC
+        mean_tpr /= self.model.n_classes
+
+        fpr["macro"] = all_fpr
+        tpr["macro"] = mean_tpr
+        roc_auc["macro"] = auc(fpr["macro"], tpr["macro"])
+
+        # Plot all ROC curves
+        plt.figure()
+        plt.plot(fpr["micro"], tpr["micro"],
+                 label=f'micro ROC ({round(roc_auc["micro"], 2)})',
+                 color='deeppink', linestyle=':', linewidth=4)
+
+        plt.plot(fpr["macro"], tpr["macro"],
+                 label=f'macro ROC({round(roc_auc["macro"], 2)})',
+                 color='navy', linestyle=':', linewidth=4)
+
+        colors = cycle(['firebrick', 'orangered', 'gold', 'olive', 'limegreen', 'aqua',
+                        'dodgerblue', 'slategrey', 'royalblue', 'indigo', 'fuchsia'], )
+
+        for i, color in zip(range(self.model.n_classes), colors):
+            plt.plot(fpr[i], tpr[i], color=color, lw=2, label=f'{class_names[i]} ({round(roc_auc[i], 2)})')
+
+        plt.plot([0, 1], [0, 1], 'k--', lw=2)
+        plt.xlim([0.0, 1.0])
+        plt.ylim([0.0, 1.05])
+        plt.xlabel('False Positive Rate')
+        plt.ylabel('True Positive Rate')
+        plt.legend(loc="lower right")
+
+        self.model.logger.log_image('ROC', image=plt.gcf(), step=self.model.current_epoch)
+        self.model.logger.log_image('ROC', image=plt.gcf(), step=self.model.current_epoch, ext='pdf')
+        plt.clf()
+
+        #######################################################################################
+        #
+        # ROC SCORE
+
+        try:
+            macro_roc_auc_ovr = roc_auc_score(y_true_one_hot, y_pred, multi_class="ovr",
+                                              average="macro")
+            summary_dict.update(macro_roc_auc_ovr=macro_roc_auc_ovr)
+        except ValueError:
+            micro_roc_auc_ovr = roc_auc_score(y_true_one_hot, y_pred, multi_class="ovr",
+                                              average="micro")
+            summary_dict.update(micro_roc_auc_ovr=micro_roc_auc_ovr)
+
+        #######################################################################################
+        #
+        # Confusion matrix
+
+        cm = confusion_matrix([class_names[x] for x in y_true], [class_names[x] for x in y_pred_max],
+                              labels=[class_names[key] for key in class_names.keys()],
+                              normalize='all')
+        disp = ConfusionMatrixDisplay(confusion_matrix=cm,
+                                      display_labels=[class_names[i] for i in range(self.model.n_classes)]
+                                      )
+        disp.plot(include_values=True)
+
+        self.model.logger.log_image('Confusion_Matrix', image=disp.figure_, step=self.model.current_epoch)
+        self.model.logger.log_image('Confusion_Matrix', image=disp.figure_, step=self.model.current_epoch, ext='pdf')
+
+        plt.close('all')
+        return summary_dict
+
+
+class BinaryScores(BaseScores):
+
+    def __init__(self, *args):
+        super(BinaryScores, self).__init__(*args)
+
+    def __call__(self, outputs):
+        summary_dict = dict()
+
+        # Additional Score like the unweighted Average Recall:
+        #########################
+        # UnweightedAverageRecall
+        y_true = torch.cat([output['batch_y'] for output in outputs]) .cpu().numpy()
+        y_pred = torch.cat([output['element_wise_recon_error'] for output in outputs]).squeeze().cpu().numpy()
+
+        # How to apply a threshold manualy
+        # y_pred = (y_pred >= 0.5).astype(np.float32)
+
+        # How to apply a threshold by IF (Isolation Forest)
+        clf = IsolationForest(random_state=self.model.seed)
+        y_score = clf.fit_predict(y_pred.reshape(-1,1))
+        y_score = (np.asarray(y_score) == -1).astype(np.float32)
+
+        uar_score = recall_score(y_true, y_score, labels=[0, 1], average='macro',
+                                 sample_weight=None, zero_division='warn')
+        summary_dict.update(dict(uar_score=uar_score))
+        #########################
+        # Precission
+        precision_score = average_precision_score(y_true, y_score)
+        summary_dict.update(dict(precision_score=precision_score))
+
+        #########################
+        # AUC
+        try:
+            auc_score = roc_auc_score(y_true=y_true, y_score=y_score)
+            summary_dict.update(dict(auc_score=auc_score))
+        except ValueError:
+            summary_dict.update(dict(auc_score=-1))
+
+        #########################
+        # pAUC
+        try:
+            pauc = roc_auc_score(y_true=y_true, y_score=y_score, max_fpr=0.15)
+            summary_dict.update(dict(pauc_score=pauc))
+        except ValueError:
+            summary_dict.update(dict(pauc_score=-1))
+
+        return summary_dict

variables.py

@@ -4,8 +4,22 @@ from argparse import Namespace
 CLEAR = 0
 MASK = 1
 
-NUM_CLASSES = 2
+# Task Options
+TASK_OPTION_multiclass = 'multiclass'
+N_CLASS_multi = 10
+multi_classes_names = ['air_conditioner', 'car_horn', 'children_playing',
+                       'dog_bar', 'drilling', 'engine_idling',
+                       'gun_shot', 'jackhammer', 'siren', 'street_music']
+multi_classes = {key: val for val, key in enumerate(multi_classes_names)}
 
 
+TASK_OPTION_binary = 'binary'
+N_CLASS_binary = 2
+binary_CLASS_clear = 0
+binary_CLASS_maske = 1
+
 # Dataset Options
-DATA_OPTIONS = Namespace(test='test', devel='devel', train='train')
+DATA_OPTION_test = 'test'
+DATA_OPTION_devel = 'devel'
+DATA_OPTION_train = 'train'
+DATA_OPTIONS = [DATA_OPTION_train, DATA_OPTION_devel, DATA_OPTION_test]