meter2800.py

from pathlib import Path
import datasets
import pandas as pd

_CITATION = """\
@misc{meter2800_dataset,
  author       = {PianistProgrammer},
  title        = {{Meter2800}: A Dataset for Music Time signature detection / Meter Classification},
  year         = {2025},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/pianistprogrammer/Meter2800}
}
"""

_DESCRIPTION = """\
Meter2800 is a dataset of 2,800 music audio samples for automatic meter classification.
Each audio file is annotated with a primary meter class label (e.g., 'two', 'three', 'four') 
and an alternative meter (numerical, e.g., 2, 3, 4, 6). 
It is split into training, validation, and test sets, each available in two class configurations:
2-class and 4-class. All audio is 16-bit WAV format.
"""

_HOMEPAGE = "https://huggingface.co/datasets/pianistprogrammer/Meter2800"
_LICENSE = "mit"

# Define the labels used in the "label" and "meter"/"alt_meter" columns
LABELS_4 = ["three", "four", "five" , "seven"]
LABELS_2 = ["three", "four"]  # Example if using a binary grouping

class Meter2800(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="4_classes", version=datasets.Version("1.0.0"),
                               description="4-class meter classification"),
        datasets.BuilderConfig(name="2_classes", version=datasets.Version("1.0.0"),
                               description="2-class meter classification"),
    ]

    def _info(self):
        label_names = LABELS_4 if self.config.name == "4_classes" else LABELS_2
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features({
                "filename": datasets.Value("string"),
                "audio": datasets.Audio(sampling_rate=None),
                "label": datasets.ClassLabel(names=label_names),
                "meter": datasets.Value("string"),
                "alt_meter": datasets.Value("string"),
            }),
            supervised_keys=("audio", "label"),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        root = Path(__file__).parent
        suffix = "4_classes" if self.config.name == "4_classes" else "2_classes"

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"csv_file": root / f"data_train_{suffix}.csv"},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={"csv_file": root / f"data_val_{suffix}.csv"},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"csv_file": root / f"data_test_{suffix}.csv"},
            ),
        ]

    def _generate_examples(self, csv_file):
        df = pd.read_csv(csv_file)
        df = df.dropna(subset=["filename", "label", "meter"]).reset_index(drop=True)

        for idx, row in df.iterrows():
            audio_path = Path(__file__).parent / row["filename"].lstrip("/")
            yield idx, {
                "filename": row["filename"],
                "audio": str(audio_path),
                "label": row["label"],
                "meter": str(row["meter"]),
                "alt_meter": str(row.get("alt_meter", row["meter"])),
            }