ocr-projects-archive / ocr-tech /tests /test_admit_page_ocr_line_pruning_threshold.py
prosperious's picture
Add 1000 files (OCR_INDRA/.gitignore ...)
f9bc0c1 verified
Raw History Blame Contribute Delete
4.36 kB
from __future__ import annotations
import hashlib
import importlib.util
import json
import sys
from argparse import Namespace
from pathlib import Path
def _load_tool():
path = Path(__file__).resolve().parents[1] / "tools" / "admit_page_ocr_line_pruning_threshold.py"
spec = importlib.util.spec_from_file_location("admit_page_ocr_line_pruning_threshold", path)
assert spec is not None
assert spec.loader is not None
module = importlib.util.module_from_spec(spec)
sys.modules[spec.name] = module
spec.loader.exec_module(module)
return module
def _sha256(path: Path) -> str:
return hashlib.sha256(path.read_bytes()).hexdigest()
def _write_json(path: Path, payload: object) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(json.dumps(payload, ensure_ascii=False) + "\n", encoding="utf-8")
def _candidate(threshold: float, cp: float, gr: float) -> dict[str, object]:
return {
"threshold": threshold,
"threshold_id": f"conf{int(round(threshold * 100)):03d}",
"documents_audit_passed": True,
"comparison_claim_safe": False,
"codepoint_absolute_improvement": cp,
"grapheme_absolute_improvement": gr,
"text_cer_codepoint_micro": 0.5 - cp,
"text_cer_grapheme_micro": 0.6 - gr,
"kept_line_count": 10,
"removed_line_count": 2,
"documents_dir": f"outputs/documents/{threshold}",
"score_dir": f"outputs/score/{threshold}",
"comparison_dir": f"outputs/comparison/{threshold}",
"documents_filter_run_sha256": "filter-run",
"documents_line_report_sha256": "line-report",
"documents_audit_sha256": "audit",
"page_eval_run_sha256": "page-run",
"metrics_sha256": "metrics",
"comparison_sha256": "comparison",
"comparison_warnings": ["not claim-safe"],
}
def _sweep(path: Path, label: str, candidates: list[dict[str, object]]) -> Path:
payload = {
"source_documents_dir": f"outputs/{label}",
"claim_safe": False,
"candidates": candidates,
}
_write_json(path, payload)
return path
def test_admit_threshold_selects_common_non_regressing_candidate(tmp_path: Path) -> None:
tool = _load_tool()
first = _sweep(
tmp_path / "first" / "ocr-line-pruning-sweep.json",
"first-page",
[_candidate(0.7, 0.02, 0.03), _candidate(0.85, 0.04, 0.04)],
)
second = _sweep(
tmp_path / "second" / "ocr-line-pruning-sweep.json",
"second-page",
[_candidate(0.7, 0.01, 0.02), _candidate(0.85, -0.20, -0.10)],
)
report = tool.admit_threshold(
Namespace(
sweep=[first, second],
out=tmp_path / "out",
min_sweeps=2,
require_common_thresholds=True,
min_codepoint_improvement=0.0,
max_codepoint_regression=0.0,
max_grapheme_regression=0.0,
)
)
assert report["selected_threshold"] == 0.7
assert report["admitted_candidate_count"] == 1
assert report["selected_candidate"]["per_page"][0]["documents_dir"] == "outputs/documents/0.7"
assert report["selected_candidate"]["per_page"][0]["page_eval_run_sha256"] == "page-run"
rejected = [row for row in report["candidates"] if row["threshold"] == 0.85][0]
assert rejected["admitted_operational"] is False
assert any("regresses" in issue for issue in rejected["issues"])
assert report["sweeps"][0]["sha256"] == _sha256(first)
def test_admit_threshold_can_require_positive_improvement_on_every_page(tmp_path: Path) -> None:
tool = _load_tool()
first = _sweep(tmp_path / "first.json", "first-page", [_candidate(0.5, 0.02, 0.01)])
second = _sweep(tmp_path / "second.json", "second-page", [_candidate(0.5, 0.0, 0.01)])
report = tool.admit_threshold(
Namespace(
sweep=[first, second],
out=tmp_path / "out",
min_sweeps=2,
require_common_thresholds=True,
min_codepoint_improvement=0.001,
max_codepoint_regression=0.0,
max_grapheme_regression=0.0,
)
)
assert report["selected_threshold"] is None
assert report["admitted_candidate_count"] == 0
assert any("minimum codepoint improvement" in issue for issue in report["candidates"][0]["issues"])