Upload dataset.py

dataset.py

@@ -0,0 +1,55 @@
+import os
+import datasets
+import soundfile as sf
+
+_CITATION = ""
+_DESCRIPTION = "Audio classification dataset with labels from folder names"
+_HOMEPAGE = ""
+_LICENSE = ""
+
+class AudioDatasetConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super(AudioDatasetConfig, self).__init__(**kwargs)
+
+class AudioDataset(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [AudioDatasetConfig(name="default", version=datasets.Version("1.0.0"))]
+
+    def _info(self):
+        labels = sorted(os.listdir(os.path.join(self.config.data_dir, "Final data")))
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features({
+                "audio": datasets.Audio(sampling_rate=16000),
+                "label": datasets.ClassLabel(names=labels)
+            }),
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = self.config.data_dir
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"data_dir": data_dir}
+            )
+        ]
+
+    def _generate_examples(self, data_dir):
+        audio_path = os.path.join(data_dir, "Final data")
+        label_list = sorted(os.listdir(audio_path))
+        id_ = 0
+        for label in label_list:
+            label_dir = os.path.join(audio_path, label)
+            if not os.path.isdir(label_dir):
+                continue
+            for fname in os.listdir(label_dir):
+                if fname.endswith(".wav"):
+                    path = os.path.join(label_dir, fname)
+                    yield id_, {
+                        "audio": path,
+                        "label": label
+                    }
+                    id_ += 1