Add 1000 files (ocr-tech/tests/test_audit_limbu_book_missed_line_review_coverage.py ...)
d82db76 verified Download ocr-tech/tools/deva_rec_torch/score.py from prosperious/ocr-projects-archive: direct link, hf CLI and curl.
- Browser
- Download file 2.03 kB
-
https://huggingface.co/prosperious/ocr-projects-archive/resolve/main/ocr-tech/tools/deva_rec_torch/score.py
- Command line
-
hf download hf://prosperious/ocr-projects-archive/ocr-tech/tools/deva_rec_torch/score.py
-
curl -L -o score.py https://huggingface.co/prosperious/ocr-projects-archive/resolve/main/ocr-tech/tools/deva_rec_torch/score.py
2.03 kB
| """Score prediction TSVs against eval labels with tools/eval_metrics.py (micro CER). | |
| Usage: score.py <out.json> system=<pred_dir> [system=<pred_dir> ...] -- name:root:labels ... | |
| Every system must have ``<pred_dir>/<name>.tsv`` for every set; a missing or short file is | |
| an error, never a silent skip. | |
| """ | |
| from __future__ import annotations | |
| import json | |
| import sys | |
| from pathlib import Path | |
| from eval_metrics import evaluate | |
| def main() -> int: | |
| out = Path(sys.argv[1]) | |
| split = sys.argv.index("--") | |
| systems = dict(arg.split("=", 1) for arg in sys.argv[2:split]) | |
| sets = [arg.split(":") for arg in sys.argv[split + 1:]] | |
| report: dict[str, dict[str, dict]] = {} | |
| for name, _root, labels in sets: | |
| gold = {rel: text for rel, text in (line.split("\t", 1) for line in Path(labels).read_text(encoding="utf-8").splitlines() if "\t" in line)} | |
| for system, pred_dir in systems.items(): | |
| pred_file = Path(pred_dir) / f"{name}.tsv" | |
| preds = {} | |
| for line in pred_file.read_text(encoding="utf-8").splitlines(): | |
| parts = line.split("\t") | |
| preds[parts[0]] = parts[1] if len(parts) > 1 else "" | |
| missing = [rel for rel in gold if rel not in preds] | |
| if missing: | |
| raise RuntimeError(f"{system}/{name}: {len(missing)} crops have no prediction (e.g. {missing[0]})") | |
| report.setdefault(name, {})[system] = evaluate([(preds[rel], text) for rel, text in gold.items()]) | |
| out.write_text(json.dumps(report, ensure_ascii=False, indent=1), encoding="utf-8") | |
| names = list(systems) | |
| print(f"{'set':<22}" + "".join(f"{s:>30}" for s in names)) | |
| for name, per_system in report.items(): | |
| cells = [f"cp {m['cer_codepoint']:.4f} gr {m['cer_grapheme']:.4f} ex {m['exact_line_acc']:.2f}" for m in (per_system[s] for s in names)] | |
| print(f"{name:<22}" + "".join(f"{c:>30}" for c in cells) + f" n={per_system[names[0]]['n_lines']}") | |
| return 0 | |
| if __name__ == "__main__": | |
| sys.exit(main()) | |