| """Schema for the gold-service export used as benchmark ground truth.""" |
|
|
| from __future__ import annotations |
|
|
| import json |
| from pathlib import Path |
| from typing import Any |
|
|
| from pydantic import BaseModel, Field, field_validator, model_validator |
|
|
|
|
| class GoldViolation(BaseModel): |
| """One independently scored localisation target.""" |
|
|
| use_case_id: str |
| use_case_name: str |
| bbox: list[float] = Field(min_length=4, max_length=4) |
| description: str | None = None |
|
|
| @field_validator("bbox") |
| @classmethod |
| def validate_bbox(cls, bbox: list[float]) -> list[float]: |
| x, y, width, height = bbox |
| if width < 0 or height < 0: |
| raise ValueError("bbox width and height must be non-negative") |
| if min(x, y) < 0 or max(x + width, y + height) > 1: |
| raise ValueError("bbox must be normalised [x, y, width, height]") |
| return bbox |
|
|
|
|
| class GoldLabel(BaseModel): |
| plausibility: int | None = Field(default=None, ge=1, le=5) |
| naturalness: int | None = Field(default=None, ge=1, le=5) |
| no_violation_detected: bool |
| violations: list[GoldViolation] = Field(default_factory=list) |
| reported: bool = False |
| report_reason: str | None = None |
|
|
| @model_validator(mode="after") |
| def validate_state(self) -> "GoldLabel": |
| if self.reported: |
| if self.violations or self.no_violation_detected: |
| raise ValueError("reported labels cannot contain evaluation targets") |
| return self |
| if self.no_violation_detected == bool(self.violations): |
| raise ValueError("label must contain violations or declare no violation, not both") |
| return self |
|
|
|
|
| class GoldRecord(BaseModel): |
| image_id: str |
| gold: GoldLabel |
| kind: str |
| base_annotator: str | None = None |
| source_annotators: list[str] = Field(default_factory=list) |
| edit_summary: Any = None |
| agreement_at_gold: dict[str, Any] | None = None |
| adjudicated_by: str | None = None |
| adjudicated_at: str | None = None |
|
|
|
|
| class GoldDataset(BaseModel): |
| """Validated top-level gold array with unique image identifiers.""" |
|
|
| records: list[GoldRecord] |
|
|
| @model_validator(mode="after") |
| def validate_unique_ids(self) -> "GoldDataset": |
| ids = [record.image_id for record in self.records] |
| if len(ids) != len(set(ids)): |
| raise ValueError("gold dataset contains duplicate image_id values") |
| return self |
|
|
| @classmethod |
| def from_file(cls, path: Path) -> "GoldDataset": |
| payload = json.loads(path.read_text(encoding="utf-8")) |
| if isinstance(payload, dict) and isinstance(payload.get("items"), list): |
| payload = payload["items"] |
| if not isinstance(payload, list): |
| raise ValueError(f"{path} must contain a gold record array") |
| return cls(records=payload) |
|
|
