Matthijs Hollemans commited on
Commit
0109099
1 Parent(s): 20c9bc6

create dataset

Browse files
Files changed (3) hide show
  1. README.md +29 -0
  2. cmu-arctic-xvectors.py +45 -0
  3. spkrec-xvect.zip +3 -0
README.md CHANGED
@@ -1,3 +1,32 @@
1
  ---
 
 
 
2
  license: mit
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ pretty_name: CMU ARCTIC X-Vectors
3
+ task_categories:
4
+ - audio
5
  license: mit
6
  ---
7
+
8
+ # Speaker embeddings extracted from CMU ARCTIC
9
+
10
+ There is one `.npy` file for each utterance in the dataset, 7931 files in total. The speaker embeddings are 512-element X-vectors.
11
+
12
+ The [CMU ARCTIC](http://www.festvox.org/cmu_arctic/) dataset divides the utterances among the following speakers:
13
+
14
+ - bdl (US male)
15
+ - slt (US female)
16
+ - jmk (Canadian male)
17
+ - awb (Scottish male)
18
+ - rms (US male)
19
+ - clb (US female)
20
+ - ksp (Indian male)
21
+
22
+ The X-vectors were extracted using [this script](https://huggingface.co/mechanicalsea/speecht5-vc/blob/main/manifest/utils/prep_cmu_arctic_spkemb.py), which uses the `speechbrain/spkrec-xvect-voxceleb` model.
23
+
24
+ Usage:
25
+
26
+ ```python
27
+ from datasets import load_dataset
28
+ embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
29
+
30
+ speaker_embeddings = embeddings_dataset[7306]["xvector"]
31
+ speaker_embeddings = torch.tensor(speaker_embeddings).unsqueeze(0)
32
+ ```
cmu-arctic-xvectors.py ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+
3
+ import os
4
+ import numpy as np
5
+ import datasets
6
+
7
+
8
+ _DATA_URL = "https://huggingface.co/datasets/Matthijs/cmu-arctic-xvectors/resolve/main/spkrec-xvect.zip"
9
+
10
+
11
+ class ArcticXvectors(datasets.GeneratorBasedBuilder):
12
+
13
+ BUILDER_CONFIGS = [
14
+ datasets.BuilderConfig(
15
+ name="default",
16
+ version=datasets.Version("0.0.1", ""),
17
+ description="",
18
+ )
19
+ ]
20
+
21
+ def _info(self):
22
+ return datasets.DatasetInfo(
23
+ features=datasets.Features(
24
+ {
25
+ "filename": datasets.Value("string"),
26
+ "xvector": datasets.Sequence(feature=datasets.Value(dtype="float32"), length=512),
27
+ }
28
+ ),
29
+ )
30
+
31
+ def _split_generators(self, dl_manager):
32
+ archive = os.path.join(dl_manager.download_and_extract(_DATA_URL), "spkrec-xvect")
33
+ return [
34
+ datasets.SplitGenerator(
35
+ name=datasets.Split.VALIDATION, gen_kwargs={"files": dl_manager.iter_files(archive)}
36
+ ),
37
+ ]
38
+
39
+ def _generate_examples(self, files):
40
+ for i, file in enumerate(sorted(files)):
41
+ if os.path.basename(file).endswith(".npy"):
42
+ yield str(i), {
43
+ "filename": os.path.basename(file)[:-4], # strip off .npy
44
+ "xvector": np.load(file),
45
+ }
spkrec-xvect.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28ea1b685a49fedce92d1af7e68b22bf511a23432bc7a13d621a4deeee9fe9a1
3
+ size 17943510