chenchaoyun commited on
Commit ·
90383e4
1
Parent(s): 71eaf61
fix
Browse files
test/remove_duplicate_celeb_images.py
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
遍历指定目录,根据文件内容(MD5)查找重复项,如果发现重复则只保留一个。
|
| 4 |
+
默认目标目录为 /opt/data/chinese_celeb_dataset,可用 --target-dir 覆盖。
|
| 5 |
+
"""
|
| 6 |
+
|
| 7 |
+
from __future__ import annotations
|
| 8 |
+
|
| 9 |
+
import argparse
|
| 10 |
+
import hashlib
|
| 11 |
+
import os
|
| 12 |
+
import sys
|
| 13 |
+
from pathlib import Path
|
| 14 |
+
from typing import Dict
|
| 15 |
+
|
| 16 |
+
DEFAULT_TARGET_DIR = Path("/opt/data/chinese_celeb_dataset")
|
| 17 |
+
CHUNK_SIZE = 4 * 1024 * 1024 # 4MB
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
def compute_md5(file_path: Path) -> str:
|
| 21 |
+
"""流式计算文件 MD5,避免一次性读入大文件。"""
|
| 22 |
+
digest = hashlib.md5()
|
| 23 |
+
with file_path.open("rb") as fh:
|
| 24 |
+
for chunk in iter(lambda: fh.read(CHUNK_SIZE), b""):
|
| 25 |
+
digest.update(chunk)
|
| 26 |
+
return digest.hexdigest()
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def deduplicate(target_dir: Path, dry_run: bool = False) -> int:
|
| 30 |
+
"""执行去重逻辑,返回删除的重复文件数量。"""
|
| 31 |
+
if not target_dir.exists():
|
| 32 |
+
print(f"[error] 目标目录不存在: {target_dir}", file=sys.stderr)
|
| 33 |
+
return 0
|
| 34 |
+
if not target_dir.is_dir():
|
| 35 |
+
print(f"[error] 目标路径不是目录: {target_dir}", file=sys.stderr)
|
| 36 |
+
return 0
|
| 37 |
+
|
| 38 |
+
md5_map: Dict[str, Path] = {}
|
| 39 |
+
removed = 0
|
| 40 |
+
scanned = 0
|
| 41 |
+
|
| 42 |
+
# 按路径排序,确保始终保留最先遍历到的文件
|
| 43 |
+
for file_path in sorted(target_dir.rglob("*")):
|
| 44 |
+
if not file_path.is_file() or file_path.is_symlink():
|
| 45 |
+
continue
|
| 46 |
+
|
| 47 |
+
scanned += 1
|
| 48 |
+
try:
|
| 49 |
+
file_md5 = compute_md5(file_path)
|
| 50 |
+
except Exception as exc:
|
| 51 |
+
print(f"[warn] 计算 MD5 失败: {file_path} -> {exc}", file=sys.stderr)
|
| 52 |
+
continue
|
| 53 |
+
|
| 54 |
+
original = md5_map.get(file_md5)
|
| 55 |
+
if original is None:
|
| 56 |
+
md5_map[file_md5] = file_path
|
| 57 |
+
continue
|
| 58 |
+
|
| 59 |
+
if dry_run:
|
| 60 |
+
print(f"[dry-run] {file_path} 与 {original} 内容相同,将被删除")
|
| 61 |
+
else:
|
| 62 |
+
try:
|
| 63 |
+
os.remove(file_path)
|
| 64 |
+
removed += 1
|
| 65 |
+
print(f"[remove] 删除重复文件: {file_path} (原始: {original})")
|
| 66 |
+
except Exception as exc:
|
| 67 |
+
print(f"[error] 删除失败: {file_path} -> {exc}", file=sys.stderr)
|
| 68 |
+
|
| 69 |
+
print(
|
| 70 |
+
f"[summary] 扫描文件: {scanned}, 保留唯一文件: {len(md5_map)}, 删除重复文件: {removed}{' (dry-run)' if dry_run else ''}"
|
| 71 |
+
)
|
| 72 |
+
return removed
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
def parse_args() -> argparse.Namespace:
|
| 76 |
+
parser = argparse.ArgumentParser(description="按 MD5 删除重复文件,仅保留一个副本。")
|
| 77 |
+
parser.add_argument(
|
| 78 |
+
"--target-dir",
|
| 79 |
+
type=Path,
|
| 80 |
+
default=DEFAULT_TARGET_DIR,
|
| 81 |
+
help=f"需要去重的目录(默认: {DEFAULT_TARGET_DIR})",
|
| 82 |
+
)
|
| 83 |
+
parser.add_argument(
|
| 84 |
+
"--dry-run",
|
| 85 |
+
action="store_true",
|
| 86 |
+
help="只输出将删除的文件,不实际删除。",
|
| 87 |
+
)
|
| 88 |
+
return parser.parse_args()
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
def main() -> int:
|
| 92 |
+
args = parse_args()
|
| 93 |
+
target_dir = args.target_dir.expanduser().resolve()
|
| 94 |
+
deduplicate(target_dir, dry_run=args.dry_run)
|
| 95 |
+
return 0
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
if __name__ == "__main__":
|
| 99 |
+
raise SystemExit(main())
|