meter2800.py

# 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 and an alternative meter.
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"

LABELS_4 = ["three", "four", "five", "seven"]
LABELS_2 = ["three", "four"]

class Meter2800Config(datasets.BuilderConfig):
    def __init__(self, name, **kwargs):
        super().__init__(name=name, version=datasets.Version("1.0.0"), **kwargs)

class Meter2800(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        Meter2800Config(name="4_classes", description="4‑class meter classification"),
        Meter2800Config(name="2_classes", description="2‑class meter classification"),
    ]
    DEFAULT_CONFIG_NAME = "4_classes"

    def _info(self):
        labels = 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=22050),
                "label": datasets.ClassLabel(names=labels),
                "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):
        csv_links = {
            split: f"https://huggingface.co/datasets/pianistprogrammer/meter2800/resolve/main/data_{split}_{self.config.name}.csv"
            for split in ["train", "val", "test"]
        }
        csv_files = dl_manager.download(csv_links)
        archive = dl_manager.download_and_extract(
            "https://huggingface.co/datasets/pianistprogrammer/meter2800/resolve/main/data.tar.gz"
        )

        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"csv_path": csv_files["train"], "root": archive}),
            datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"csv_path": csv_files["val"], "root": archive}),
            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"csv_path": csv_files["test"], "root": archive}),
        ]

    def _generate_examples(self, csv_path, root):
        df = pd.read_csv(csv_path).dropna(subset=["filename", "label", "meter"]).reset_index(drop=True)
        for idx, row in df.iterrows():
            rel = row["filename"].lstrip("/")  # ensure relative path, not absolute
            audio_path = Path(root) / rel
            if not audio_path.is_file():
                raise FileNotFoundError(f"Missing audio file: {audio_path}")
            yield idx, {
                "filename": rel,
                "audio": str(audio_path),
                "label": row["label"],
                "meter": str(row["meter"]),
                "alt_meter": str(row.get("alt_meter", row["meter"])),
            }