Add auditable SAM roof label refinement

scripts/build_regional_yolo_dataset.py

@@ -0,0 +1,77 @@
+#!/usr/bin/env python3
+"""Create a checksummed regional YOLO view without copying protected data."""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as stream:
+        for chunk in iter(lambda: stream.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def select_paths(summary: dict[str, Any], manifest: dict[str, Any], region: str) -> tuple[list[str], list[str]]:
+    samples = {item["sample_slug"]: item for item in manifest["samples"]}
+    train: list[str] = []
+    val: list[str] = []
+    for tile in summary["tiles"]:
+        sample = samples[tile["sample_slug"]]
+        if sample["region"] != region or not tile.get("kept", True):
+            continue
+        if sample["split"] == "train" and tile["split"] == "train":
+            train.append(tile["image_path"])
+        elif sample["split"] == "val" and tile["split"] == "val":
+            val.append(tile["image_path"])
+    if not train or not val:
+        raise ValueError(f"Region {region!r} must contain train and validation images")
+    return sorted(train), sorted(val)
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--summary", type=Path, required=True)
+    parser.add_argument("--corpus-manifest", type=Path, required=True)
+    parser.add_argument("--region", required=True)
+    parser.add_argument("--output-dir", type=Path, required=True)
+    args = parser.parse_args()
+    summary = json.loads(args.summary.read_text(encoding="utf-8"))
+    manifest = json.loads(args.corpus_manifest.read_text(encoding="utf-8"))
+    train, val = select_paths(summary, manifest, args.region)
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    train_list = args.output_dir / "train.txt"
+    val_list = args.output_dir / "val.txt"
+    train_list.write_text("\n".join(train) + "\n", encoding="utf-8")
+    val_list.write_text("\n".join(val) + "\n", encoding="utf-8")
+    dataset_yaml = args.output_dir / "dataset.yaml"
+    dataset_yaml.write_text(
+        f"path: {args.output_dir}\ntrain: {train_list}\nval: {val_list}\nnames:\n  0: building\n",
+        encoding="utf-8",
+    )
+    evidence = {
+        "schema_version": 1,
+        "status": "ok",
+        "region": args.region,
+        "train_image_count": len(train),
+        "validation_image_count": len(val),
+        "summary": str(args.summary),
+        "summary_sha256": sha256(args.summary),
+        "corpus_manifest": str(args.corpus_manifest),
+        "corpus_manifest_sha256": sha256(args.corpus_manifest),
+        "dataset_yaml": str(dataset_yaml),
+        "protected_splits_in_training": [],
+    }
+    (args.output_dir / "regional-dataset.json").write_text(json.dumps(evidence, indent=2), encoding="utf-8")
+    print(json.dumps(evidence, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())