Upload 11 files

README.md

@@ -0,0 +1,47 @@
+---
+tags:
+- audio
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-00000-of-00001.parquet
+---
+
+# Background Noise Dataset
+
+This dataset contains **3** audio recordings of **2** different background noise classes.
+
+## Dataset Statistics
+
+- **Total Audio Files**: 3
+- **Total Classes**: 2
+- **Format**: WAV (converted to Parquet for Hugging Face Datasets)
+
+## Classes and Descriptions
+
+The dataset covers the following background noises:
+
+| Label | Description |
+| :--- | :--- |
+| `fan_noise` | Fan noise background |
+| `white_noise` | White noise background |
+
+## Structure
+
+The dataset is organized in a folder structure where each subdirectory corresponds to a class label.
+A `metadata.csv` file provides the file paths, labels, and descriptions.
+
+### Columns
+
+- `audio`: Path to the audio file.
+- `label`: The category label of the background noise.
+- `description`: A text description of the background noise.
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("sdialog/background")
+```

convert_to_parquet.py

@@ -0,0 +1,64 @@
+import os
+import csv
+from datasets import Dataset, Audio, Features, Value, Version
+
+# Point to the current directory
+dataset_path = os.path.dirname(os.path.abspath(__file__))
+
+# Dataset metadata
+VERSION = Version("1.0.0")
+DESCRIPTION = "Background noise dataset."
+HOMEPAGE = "https://huggingface.co/datasets/sdialog/background"
+LICENSE = "CC BY-NC-SA 4.0"
+
+# Define features
+features = Features({
+    "audio": Audio(sampling_rate=16000),
+    "file_path": Value("string"),
+    "label": Value("string"),
+    "description": Value("string"),
+})
+
+
+def dataset_generator():
+    csv_path = os.path.join(dataset_path, "metadata.csv")
+    if not os.path.exists(csv_path):
+        raise FileNotFoundError(
+            f"{csv_path} not found. Run create_metadata.py first."
+        )
+
+    with open(csv_path, "r", encoding="utf-8") as f:
+        reader = csv.DictReader(f)
+        for row in reader:
+            audio_path = row["audio"]
+            # Ensure the path is absolute for reading
+            # audio_path is relative to the dataset root
+            full_path = os.path.join(dataset_path, audio_path)
+
+            # Read audio bytes to ensure embedding
+            with open(full_path, "rb") as audio_file:
+                audio_bytes = audio_file.read()
+
+            yield {
+                "audio": {"path": None, "bytes": audio_bytes},
+                "file_path": audio_path,
+                "label": row["label"],
+                "description": row["description"]
+            }
+
+
+print("Creating dataset from generator (embedding bytes)...")
+ds = Dataset.from_generator(dataset_generator, features=features)
+
+# Set dataset info
+ds.info.version = VERSION
+ds.info.description = DESCRIPTION
+ds.info.homepage = HOMEPAGE
+ds.info.license = LICENSE
+
+print("Saving to Parquet...")
+output_path = os.path.join(dataset_path, "data/train-00000-of-00001.parquet")
+os.makedirs(os.path.dirname(output_path), exist_ok=True)
+ds.to_parquet(output_path)
+
+print(f"Saved to {output_path}")

create_metadata.py

@@ -0,0 +1,58 @@
+import os
+import csv
+
+# Define paths
+base_path = os.path.dirname(os.path.abspath(__file__))
+# The audio folders are directly in the base path for this dataset
+train_dir = base_path
+tags_path = os.path.join(base_path, "tags_labels.csv")
+output_csv = os.path.join(base_path, "metadata.csv")
+
+# Load tags/descriptions
+labels_desc = {}
+if os.path.exists(tags_path):
+    with open(tags_path, 'r', encoding='utf-8') as file_tags:
+        for line in file_tags:
+            parts = line.strip().split(';')
+            if len(parts) >= 2:
+                tag_name = parts[0].strip()
+                tag_description = parts[1].strip()
+                labels_desc[tag_name] = tag_description
+else:
+    print(f"Warning: {tags_path} not found.")
+
+# Prepare CSV data
+csv_data = []
+header = ["audio", "label", "description"]
+
+if os.path.exists(train_dir):
+    # Iterate over label directories
+    for label in sorted(os.listdir(train_dir)):
+        label_dir = os.path.join(train_dir, label)
+        
+        # Skip hidden files, non-directories, and special folders/files
+        if not os.path.isdir(label_dir) or label.startswith('.'):
+            continue
+        
+        # Only process directories that are likely audio labels (simple heuristic or check against tags)
+        # Here we just check if it's not the 'data' output folder or hidden
+        if label == "data":
+            continue
+
+        description = labels_desc.get(label, "XXX")
+        
+        # Iterate over audio files in the label directory
+        for audio_file in sorted(os.listdir(label_dir)):
+            if audio_file.lower().endswith(('.wav', '.mp3', '.ogg', '.flac')):
+                # Create relative path: label/file.wav (relative to metadata.csv location)
+                relative_path = os.path.join(label, audio_file)
+                
+                csv_data.append([relative_path, label, description])
+
+# Write to CSV
+with open(output_csv, "w", newline="", encoding="utf-8") as f:
+    writer = csv.writer(f)
+    writer.writerow(header)
+    writer.writerows(csv_data)
+
+print(f"Successfully created {output_csv} with {len(csv_data)} entries.")

data/train-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a004d9643330d231dc5d9f33f68bd1231f1cab7389724fed1c475bb1fe712d5a
+size 56822341

fan_noise/210098__yuval__room-big-fan.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91411e5b157cf23d925abac6bd38ee36c5701d3a4011a9a8f9d383c5ad223347
+size 24617496

fan_noise/674563__klankbeeld__room-tone-fan-801am-220813_0497.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c735017ce0aafc2c4b5fd3b13dc3de31f52b3b6caae36b429b87ca8e00bc9011
+size 29840984

generate_data.sh

@@ -0,0 +1 @@
+python create_metadata.py && python convert_to_parquet.py

metadata.csv

@@ -0,0 +1,4 @@
+audio,label,description
+fan_noise/210098__yuval__room-big-fan.wav,fan_noise,Fan noise background
+fan_noise/674563__klankbeeld__room-tone-fan-801am-220813_0497.wav,fan_noise,Fan noise background
+white_noise/white_zero_zero.wav,white_noise,White noise background

tags_labels.csv

@@ -0,0 +1,2 @@
+fan_noise;Fan noise background
+white_noise;White noise background

test_dataset.py

@@ -0,0 +1,26 @@
+from datasets import load_dataset
+import os
+
+# Point to the parquet file generated
+dataset_path = os.path.join(os.path.dirname(__file__), "data/train-00000-of-00001.parquet")
+
+if not os.path.exists(dataset_path):
+    print(f"Dataset file not found at {dataset_path}")
+    print("Please run convert_to_parquet.py first.")
+    exit(1)
+
+print(f"Loading background dataset from {dataset_path}...")
+ds = load_dataset("parquet", data_files={"train": dataset_path}, split="train")
+
+print("\nDataset Info:")
+print(ds)
+
+print("\nFeatures:")
+print(ds.features)
+
+print("\nFirst Example:")
+item = ds[0]
+print(item)
+
+print("\nChecking file_path:")
+print(f"File Path: {item['file_path']}")

white_noise/white_zero_zero.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cb253ff26c63417a0fa82ea52af5ce897d9275c553477087ece7fb1e20eefca
+size 9024188