Upload extract_images.py with huggingface_hub

extract_images.py

@@ -0,0 +1,41 @@
+"""Extract SoloFace2 image archives.
+
+Download all images_part_*.tar.gz files to this directory and run:
+    python extract_images.py
+
+This will extract all 311,717 images into the images/ directory.
+Already-extracted images are skipped (safe to re-run).
+"""
+import tarfile, sys, os
+from pathlib import Path
+
+ARCHIVE_PREFIX = 'images_part_'
+OUTPUT_DIR = Path(__file__).parent / 'images'
+OUTPUT_DIR.mkdir(exist_ok=True)
+
+# Find all archive files
+archives = sorted(Path(__file__).parent.glob(f'{ARCHIVE_PREFIX}*.tar.gz'))
+if not archives:
+    print('ERROR: No images_part_*.tar.gz files found!')
+    print('Download all archive files to this directory first.')
+    sys.exit(1)
+
+print(f'Found {len(archives)} archive(s):')
+for a in archives:
+    print(f'  {a.name} ({a.stat().st_size / 1e6:.0f} MB)')
+
+total_extracted = 0
+for archive in archives:
+    print(f'\nExtracting {archive.name}...')
+    with tarfile.open(archive, 'r:gz') as tar:
+        members = tar.getmembers()
+        for i, member in enumerate(members):
+            target = OUTPUT_DIR / member.name.split('/')[-1]
+            if target.exists():
+                continue  # Skip already-extracted
+            tar.extract(member, path=OUTPUT_DIR.parent, filter='data')
+            if (i + 1) % 10000 == 0:
+                print(f'  {i+1:,}/{len(members):,}')
+    total_extracted += len(members)
+
+print(f'\nDone! {total_extracted:,} images in {OUTPUT_DIR}/')