Upload ATD2.py

ATD2.py

@@ -0,0 +1,59 @@
+import os
+import pandas as pd
+from datasets import GeneratorBasedBuilder, DatasetInfo, Split, SplitGenerator, Features, Value, Audio
+
+# Code on HF used for dataset gen
+
+class ATD2(GeneratorBasedBuilder):
+
+    def _info(self):
+        return DatasetInfo(
+            description="Audio classification dataset with 10 folds and metadata.",
+            features=Features({
+                "audio": Audio(), 
+                "label": Value("string"),
+                "audio_path": Value("string"),
+                "fold": Value("int32")
+            }),
+            supervised_keys=("audio", "label"),
+        )
+
+    def _split_generators(self, dl_manager):
+        print(" _split_generators running...")
+
+        extracted_dir = dl_manager.download_and_extract(
+            "https://huggingface.co/datasets/mmzoellner/ATD2/resolve/main/folds.tar"
+        )
+        metadata_path = dl_manager.download(
+            "https://huggingface.co/datasets/mmzoellner/ATD2/resolve/main/metadata.csv"
+        )
+
+        return [
+            SplitGenerator(
+                name=Split.TRAIN,
+                gen_kwargs={
+                    "archive_path": extracted_dir,
+                    "metadata_path": metadata_path
+                }
+            )
+        ]
+
+    def _generate_examples(self, archive_path, metadata_path):
+        print(" _generate_examples is running...")
+        df = pd.read_csv(metadata_path)
+    
+        for idx, row in df.iterrows():
+            corrected_path = row["audio_path"].replace("audio/", "")
+            full_path = os.path.join(archive_path, corrected_path)
+        
+            if not os.path.isfile(full_path):
+                print(f"Missing: {full_path}")
+                continue
+        
+            yield idx, {
+                "audio": full_path,
+                "label": row["label"],
+                "audio_path": corrected_path,
+                "fold": int(row["fold"]),
+            }
+        print("done.")