gengxin

2026-07-18 09:13:04 +08:00
parent 271f0c4efd
commit 8bbc9da1e5
62 changed files with 2048 additions and 9 deletions
@@ -0,0 +1,116 @@
+import json
+import math
+from pathlib import Path
+
+from PIL import Image, ImageDraw, ImageFont, ImageOps
+
+
+ROOT = Path(__file__).resolve().parents[1]
+BASE = ROOT / ".tmp" / "gzhh-images"
+RAW = BASE / "raw"
+MANIFEST = BASE / "manifest.json"
+AUDIT = BASE / "audit.json"
+SHEETS = BASE / "contact-sheets"
+
+CELL_WIDTH = 280
+CELL_HEIGHT = 220
+COLS = 5
+ROWS = 4
+PER_SHEET = COLS * ROWS
+
+
+def font(size: int):
+    paths = [
+        Path("C:/Windows/Fonts/msyh.ttc"),
+        Path("C:/Windows/Fonts/simhei.ttf"),
+    ]
+    for path in paths:
+        if path.exists():
+            return ImageFont.truetype(str(path), size)
+    return ImageFont.load_default()
+
+
+small_font = font(14)
+label_font = font(16)
+title_font = font(22)
+
+
+manifest = json.loads(MANIFEST.read_text(encoding="utf-8"))
+unique_entries = []
+seen_files = set()
+
+for entry in manifest:
+    file_name = entry.get("file")
+    if entry.get("status") != "ok" or not file_name or file_name in seen_files:
+        continue
+    seen_files.add(file_name)
+    image_path = RAW / file_name
+    record = {
+        "id": len(unique_entries) + 1,
+        "file": file_name,
+        "path": image_path.relative_to(ROOT).as_posix(),
+        "sources": entry.get("sources", []),
+        "alts": entry.get("alts", []),
+        "bytes": entry.get("bytes", 0),
+        "hash": entry.get("hash", ""),
+    }
+    try:
+        with Image.open(image_path) as image:
+            record["width"], record["height"] = image.size
+            record["format"] = image.format
+            preview = image.convert("RGB")
+            preview.thumbnail((128, 128))
+            record["entropy"] = round(preview.convert("L").entropy(), 3)
+            record["candidate"] = (
+                record["width"] * record["height"] >= 200_000
+                and min(record["width"], record["height"]) >= 280
+                and 0.32 <= record["width"] / record["height"] <= 3.2
+                and record["entropy"] >= 3.2
+            )
+    except Exception as error:
+        record["error"] = str(error)
+        record["candidate"] = False
+    unique_entries.append(record)
+
+AUDIT.write_text(json.dumps(unique_entries, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+SHEETS.mkdir(parents=True, exist_ok=True)
+
+candidates = [item for item in unique_entries if item.get("candidate")]
+
+for sheet_index in range(math.ceil(len(candidates) / PER_SHEET)):
+    page_items = candidates[sheet_index * PER_SHEET : (sheet_index + 1) * PER_SHEET]
+    sheet = Image.new("RGB", (CELL_WIDTH * COLS, 48 + CELL_HEIGHT * ROWS), "#f7f3e8")
+    draw = ImageDraw.Draw(sheet)
+    draw.text(
+        (18, 10),
+        f"公众号图片候选 {sheet_index + 1}/{math.ceil(len(candidates) / PER_SHEET)} · {len(candidates)} 张",
+        fill="#1f2b2a",
+        font=title_font,
+    )
+    for item_index, item in enumerate(page_items):
+        col = item_index % COLS
+        row = item_index // COLS
+        x = col * CELL_WIDTH
+        y = 48 + row * CELL_HEIGHT
+        draw.rectangle((x + 5, y + 5, x + CELL_WIDTH - 5, y + CELL_HEIGHT - 5), fill="#ffffff", outline="#d6d7cf")
+        try:
+            with Image.open(ROOT / item["path"]) as source:
+                preview = ImageOps.contain(source.convert("RGB"), (CELL_WIDTH - 24, 154))
+                px = x + (CELL_WIDTH - preview.width) // 2
+                py = y + 12 + (154 - preview.height) // 2
+                sheet.paste(preview, (px, py))
+        except Exception:
+            pass
+        source_name = Path(item["sources"][0]).stem if item["sources"] else "未知来源"
+        source_name = source_name[:20]
+        draw.text((x + 12, y + 170), f"#{item['id']:03d}  {item['width']}×{item['height']}", fill="#376d5a", font=label_font)
+        draw.text((x + 12, y + 192), source_name, fill="#596563", font=small_font)
+    sheet.save(SHEETS / f"sheet-{sheet_index + 1:02d}.jpg", quality=90)
+
+summary = {
+    "unique": len(unique_entries),
+    "candidates": len(candidates),
+    "rejected_by_shape_or_size": len(unique_entries) - len(candidates),
+    "sheets": math.ceil(len(candidates) / PER_SHEET),
+}
+print(json.dumps(summary, ensure_ascii=False))