First commit

configs/parameters.yaml

+# PATHS
+wav_path: '../data/wavs/'
+embedding_path: '../data/embeddings/'
+
+# PARAMETERS
+sampling_rate: 16000
+batch_size: 16
+learning_rate: 0.001
+
+# Types of embeddings supported: 'ecapa_tdnn' or 'x-vector'
+embedding_type: ecapa_tdnn

data/wavs/test.txt

+sjfnjsdnf

dataloader/embedding_extract.py

+import torchaudio
+import torch
+import sys
+import os
+import yaml
+from speechbrain.pretrained import EncoderClassifier
+from tqdm import tqdm
+
+
+# LOADS CONFIGURATION .YAML FILE
+def load_config(config_path):
+    try:
+        with open(config_path, 'r') as file:
+            config = yaml.safe_load(file)
+        return config
+    except Exception as e:
+        print('Error reading the config file')
+
+# RETURNS THE EMBEDDING EXTRACTO
+# TWO SUPPORTED TYPES: X-VECTORS AND ECAPA_TDNN
+def embedding_type():
+    if config['embedding_type'] == 'x-vector':
+        classifier = EncoderClassifier.from_hparams(source=\
+                "speechbrain/spkrec-xvect-voxceleb", savedir=\
+                "pretrained_models/spkrec-xvect-voxceleb")
+        extension = 'xvec'
+    elif config['embedding_type'] == 'ecapa_tdnn':
+        classifier = EncoderClassifier.from_hparams(source=\
+                "speechbrain/spkrec-ecapa-voxceleb", savedir=\
+                "pretrained_models/spkrec-ecapatdnn-voxceleb")
+        extension = 'ecapa'
+    else:
+        print('Unknown embedding type')
+        sys.exit()
+
+    return classifier, extension
+
+# EXTRACTS THE SPEAKER EMBEDDING BASED ON THE USED CLASSIFIER
+# CLASSIFIERS OBTAINED FROM HUGGINGFACE
+def extractor(wavfile, classifier):
+    
+    signal, fs =torchaudio.load(wavfile)
+    embedding = classifier.encode_batch(signal)
+
+    return embedding
+
+
+
+
+if __name__=="__main__":
+
+    config_path = '../configs/parameters.yaml'
+    config = load_config(config_path)
+
+    extraction_files = os.listdir(config['wav_path'])
+
+    classifier, ext = embedding_type()
+    saving_path = config['embedding_path']+ext+'/'
+    
+    # CREATES SPECIFIC EMBEDDING SUBDIR IF IT DOESN'T EXIST
+    if not os.path.exists(saving_path):
+        os.makedirs(saving_path)
+
+
+    print("\nExtracting speaker embeddings...")
+    with tqdm(total=len(extraction_files)) as pbar:
+        for files in os.listdir(config['wav_path']):
+            if files.endswith('.wav'):
+                embedding = extractor(config['wav_path'] + files, classifier)
+                buf = os.path.splitext(files)
+                new_f = buf[0]+'_'+ext+'_emb.pickle'
+            
+                torch.save(embedding, saving_path+new_f)
+                pbar.update(1)
+            else:
+                pbar.update(1)
+                continue
+
+
+

dataloader/pretrained_models/spkrec-ecapatdnn-voxceleb/classifier.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/5c0be3875fda05e81f3c004ed8c7c06be308de1e/classifier.ckpt
\ No newline at end of file

dataloader/pretrained_models/spkrec-ecapatdnn-voxceleb/embedding_model.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/5c0be3875fda05e81f3c004ed8c7c06be308de1e/embedding_model.ckpt
\ No newline at end of file

dataloader/pretrained_models/spkrec-ecapatdnn-voxceleb/hyperparams.yaml

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/5c0be3875fda05e81f3c004ed8c7c06be308de1e/hyperparams.yaml
\ No newline at end of file

dataloader/pretrained_models/spkrec-ecapatdnn-voxceleb/label_encoder.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/5c0be3875fda05e81f3c004ed8c7c06be308de1e/label_encoder.txt
\ No newline at end of file

dataloader/pretrained_models/spkrec-ecapatdnn-voxceleb/mean_var_norm_emb.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/5c0be3875fda05e81f3c004ed8c7c06be308de1e/mean_var_norm_emb.ckpt
\ No newline at end of file

dataloader/pretrained_models/spkrec-xvect-voxceleb/classifier.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-xvect-voxceleb/snapshots/e2cc27f853f99bd5d539432f0cba3f124c059f71/classifier.ckpt
\ No newline at end of file

dataloader/pretrained_models/spkrec-xvect-voxceleb/embedding_model.ckpt

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-xvect-voxceleb/snapshots/e2cc27f853f99bd5d539432f0cba3f124c059f71/embedding_model.ckpt
\ No newline at end of file

dataloader/pretrained_models/spkrec-xvect-voxceleb/hyperparams.yaml

+/home/squintas/.cache/huggingface/hub/models--speechbrain--spkrec-xvect-voxceleb/snapshots/e2cc27f853f99bd5d539432f0cba3f124c059f71/hyperparams.yaml
\ No newline at end of file