from __future__ import annotations import csv from dataclasses import dataclass from pathlib import Path import numpy as np @dataclass(frozen=True) class Dataset: image_paths: tuple[Path, ...] file_names: np.ndarray labels: np.ndarray classes: tuple[str, ...] def read_rows(path: Path) -> list[dict[str, str]]: with path.open(newline="", encoding="utf-8-sig") as handle: return list(csv.DictReader(handle)) def load_dataset(root: Path) -> Dataset: rows = read_rows(root / "human-tagged.txt.txt") classes = tuple(sorted({label for row in rows for label in row["characters"].split("|")})) class_to_index = {name: index for index, name in enumerate(classes)} image_paths: list[Path] = [] labels: list[np.ndarray] = [] for row in rows: path = root / row["file_name"] if not path.is_file(): raise FileNotFoundError(path) target = np.zeros(len(classes), dtype=np.uint8) for name in row["characters"].split("|"): target[class_to_index[name]] = 1 image_paths.append(path) labels.append(target) return Dataset( image_paths=tuple(image_paths), file_names=np.asarray([path.name for path in image_paths]), labels=np.stack(labels), classes=classes, )