add segmentation (without saving file on disk)

voxpopuli.py

@@ -1,5 +1,9 @@
+from collections import defaultdict
 import os
 import glob
+import csv
+from pathlib import Path
+from tqdm.auto import tqdm
 
 import datasets
 
@@ -14,6 +18,7 @@ _LANGUAGES_V2 = [f"{x}_v2" for x in _LANGUAGES]
 
 _YEARS = list(range(2009, 2020 + 1))
 
+# unnecessary
 _CONFIG_TO_LANGS = {
     "400k": _LANGUAGES,
     "100k": _LANGUAGES,
@@ -33,7 +38,20 @@ _BASE_URL = "https://dl.fbaipublicfiles.com/voxpopuli/"
 
 _DATA_URL = _BASE_URL + "audios/{lang}_{year}.tar"
 
-_META_URL = _BASE_URL + "https://dl.fbaipublicfiles.com/voxpopuli/annotations/unlabelled_v2.tsv.gz"
+_META_URL = _BASE_URL + "annotations/unlabelled_v2.tsv.gz"
+
+
+class VoxpopuliConfig(datasets.BuilderConfig):
+    """BuilderConfig for VoxPopuli."""
+
+    def __init__(self, name, **kwargs):
+        """
+        Args:
+          name: `string`, name of dataset config
+          **kwargs: keyword arguments forwarded to super.
+        """
+        super().__init__(name=name, **kwargs)
+        self.languages = [name] if name in _LANGUAGES else _LANGUAGES
 
 
 class Voxpopuli(datasets.GeneratorBasedBuilder):
@@ -41,7 +59,7 @@ class Voxpopuli(datasets.GeneratorBasedBuilder):
 
     VERSION = datasets.Version("1.0.0")  # TODO ??
     BUILDER_CONFIGS = [
-        datasets.BuilderConfig(
+        VoxpopuliConfig(
             name=name,
             # version=VERSION,
             description="",  # TODO
@@ -58,6 +76,7 @@ class Voxpopuli(datasets.GeneratorBasedBuilder):
                 "language": datasets.ClassLabel(names=_LANGUAGES),
                 "year": datasets.Value("int16"),
                 "audio": datasets.Audio(sampling_rate=16_000),
+                "segment_id": datasets.Value("int16"),
             }
         )
         return datasets.DatasetInfo(
@@ -68,38 +87,75 @@ class Voxpopuli(datasets.GeneratorBasedBuilder):
             # citation=_CITATION,
         )
 
-    def _split_generators(self, dl_manager):
-        # dl_manager.download_config.num_proc = len(_VOXPOPULI_AUDIO_URLS)  # TODO
+    def _read_metadata(self, metadata_path):
+        # TODO: check for predicate??
+        #  @ https://github.com/facebookresearch/voxpopuli/blob/main/voxpopuli/get_unlabelled_data.py#L34
+        metadata = defaultdict(list)
 
-        # metadata_path = dl_manager.download_and_extract(_META_URL)
+        with open(metadata_path, encoding="utf-8") as csv_file:
+            csv_reader = csv.reader(csv_file, delimiter="\t")
+            for i, row in tqdm(enumerate(csv_reader)):
+                if i == 0:
+                    continue
+                audio_id, segment_id, start, end = row
+                event_id, lang = audio_id.rsplit("_", 1)[-2:]
+                if lang in self.languages:
+                # if lang in ["hr", "et"]:
+                    metadata[audio_id].append((float(start), float(end)))
 
-        languages = [self.config.name] if self.config.name in _LANGUAGES else _LANGUAGES
-        years = _CONFIG_TO_YEARS[self.config.name]
-        # urls = [_DATA_URL.format(lang=language, year=year) for language in ["hr", "et"] for year in [2020]]
-        urls = [_DATA_URL.format(lang=language, year=year) for language in languages for year in years]
+        return metadata
+
+    def _split_generators(self, dl_manager):
+        metadata_path = dl_manager.download_and_extract(_META_URL)
 
-        langs_data_dirs = dl_manager.download_and_extract(urls)
-        print(langs_data_dirs)
-        print(glob.glob(f"{langs_data_dirs[0]}/**/*.ogg", recursive=True))
+        years = _CONFIG_TO_YEARS[self.config.name]
+        # urls = [_DATA_URL.format(lang=language, year=year) for language in ["hr", "et"] for year in [2020]]  # , "et"]
+        urls = [_DATA_URL.format(lang=language, year=year) for language in self.languages for year in years]
+        dl_manager.download_config.num_proc = len(urls)
+        data_dirs = dl_manager.download_and_extract(urls)
 
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "data_dirs": langs_data_dirs,
+                    "data_dirs": data_dirs,
+                    "metadata_path": metadata_path,
                 }
             ),
         ]
 
-    def _generate_examples(self, data_dirs):
+    def _generate_examples(self, data_dirs, metadata_path):
+        try:
+            import torch
+            import torchaudio
+        except ImportError as e:
+            raise ValueError(
+                "Loading voxpopuli requires `torchaudio` to be installed."
+                "You can install torchaudio with `pip install torchaudio`." + e
+            )
+
+        metadata = self._read_metadata(metadata_path)
+
         for data_dir in data_dirs:
-            for file in glob.glob(f"{data_dir}/**/*.ogg", recursive=True):
+            for file in glob.glob(f"{data_dir}/**/*.ogg", recursive=True)[:5]:
                 path_components = file.split(os.sep)
-                language, year = path_components[-3:-1]
-                with open(file) as f:
-                    yield file, {
+                language, year, audio_filename = path_components[-3:]
+                audio_id, _ = os.path.splitext(audio_filename)
+                timestamps = metadata[audio_id]
+
+                waveform, sr = torchaudio.load(file)
+                duration = waveform.size(1)
+
+                for segment_id, (start, stop) in enumerate(timestamps):
+                    segment = waveform[:, int(start * sr): min(int(stop * sr), duration)]
+
+                    yield f"{audio_filename}_{segment_id}", {
                         "path": file,
                         "language": language,
                         "year": year,
-                        "audio": {"path": file}
+                        "audio": {
+                            "array": segment[0],  # segment is a 2-dim array
+                            "sampling_rate": 16_000
+                        },
+                        "segment_id": segment_id,
                     }