manual submit

2026-03-07 21:49:16 +07:00
parent 1748cbf8d2
commit 6004b000a7
39 changed files with 5794 additions and 614 deletions
--- a/sync_faces.py
+++ b/sync_faces.py
@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+"""
+sync_faces.py
+=============
+
+One-time backfill script: scan existing input_images and generation_files
+for faces and store detections in faces.db.
+
+Usage:
+    python sync_faces.py [--dry-run] [--input-only] [--output-only]
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import logging
+import sqlite3
+
+logging.basicConfig(level=logging.INFO, format="%(levelname)s: %(message)s")
+logger = logging.getLogger(__name__)
+
+
+async def main(
+    dry_run: bool, input_only: bool, output_only: bool,
+    cluster: bool, cluster_threshold: float,
+) -> None:
+    import face_db
+    from face_service import get_face_service
+
+    face_db.init_db()
+    svc = get_face_service()
+
+    if not svc.available:
+        logger.error(
+            "insightface is not available. "
+            "Install: pip install insightface onnxruntime opencv-python"
+        )
+        return
+
+    import generation_db
+    import input_image_db
+
+    total_faces = 0
+    total_matched = 0
+    total_unidentified = 0
+
+    # Scan input images
+    if not output_only:
+        logger.info("Scanning input images…")
+        conn = sqlite3.connect(str(input_image_db.DB_PATH), check_same_thread=False)
+        conn.row_factory = sqlite3.Row
+        rows = conn.execute(
+            "SELECT id, image_data FROM input_images WHERE image_data IS NOT NULL"
+        ).fetchall()
+        conn.close()
+
+        for row in rows:
+            row_id = row["id"]
+            image_bytes = bytes(row["image_data"])
+            logger.info("  input image id=%d (%d bytes)", row_id, len(image_bytes))
+            if not dry_run:
+                try:
+                    results = await svc.scan_input_image(row_id, image_bytes)
+                    for r in results:
+                        total_faces += 1
+                        if r.matched_person_id is not None:
+                            total_matched += 1
+                        else:
+                            total_unidentified += 1
+                except Exception as exc:
+                    logger.warning("  Failed for input id=%d: %s", row_id, exc)
+
+    # Scan generated output files
+    if not input_only:
+        logger.info("Scanning generation output files…")
+        conn = sqlite3.connect(str(generation_db._DB_PATH), check_same_thread=False)
+        conn.row_factory = sqlite3.Row
+        rows = conn.execute(
+            "SELECT id, file_data, mime_type FROM generation_files"
+        ).fetchall()
+        conn.close()
+
+        for row in rows:
+            file_id = row["id"]
+            file_data = bytes(row["file_data"])
+            mime_type = row["mime_type"] or ""
+            logger.info(
+                "  output file id=%d mime=%s (%d bytes)", file_id, mime_type, len(file_data)
+            )
+            if not dry_run:
+                try:
+                    if mime_type.startswith("image/"):
+                        results = await svc.scan_output_image(file_id, file_data)
+                        total_faces += len(results)
+                        total_matched += sum(1 for r in results if r.matched_person_id is not None)
+                    elif mime_type.startswith("video/"):
+                        results = await svc.scan_video(file_id, file_data)
+                        total_faces += len(results)
+                        total_matched += sum(1 for r in results if r.matched_person_id is not None)
+                except Exception as exc:
+                    logger.warning("  Failed for output id=%d: %s", file_id, exc)
+
+    if dry_run:
+        logger.info("Dry run — no data written.")
+    else:
+        logger.info(
+            "Done. %d faces detected, %d matched to known persons, %d unidentified",
+            total_faces,
+            total_matched,
+            total_unidentified,
+        )
+        if cluster:
+            logger.info("Clustering unidentified faces (threshold=%.2f)…", cluster_threshold)
+            groups = await svc.cluster_unidentified_faces(cluster_threshold)
+            logger.info("Clustering: %d groups created", len(groups))
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Backfill face detections for existing media")
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--input-only", action="store_true")
+    parser.add_argument("--output-only", action="store_true")
+    parser.add_argument("--cluster", action="store_true", help="Run auto-clustering after scanning")
+    parser.add_argument("--cluster-threshold", type=float, default=0.45, metavar="T",
+                        help="Cosine similarity threshold for clustering (default: 0.45)")
+    args = parser.parse_args()
+    asyncio.run(main(
+        args.dry_run, args.input_only, args.output_only,
+        args.cluster, args.cluster_threshold,
+    ))