95 lines
4.3 KiB
Python
95 lines
4.3 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
从 Wikimedia Commons 抓真实图片（公共领域 / CC），供 huashu-design「内容型设计取真图」用（Phase 3.5）。
|
||
|
||
为什么有这个脚本：内容型设计（鹦鹉/咖啡/马来西亚…）必须用真图，不能 CSS 色块糊弄。
|
||
每次让模型现写抓图逻辑既慢又容易漏坑（忘清代理→TLS 炸 / 忘合规 UA→429）。这里固化好，下次只改关键词。
|
||
|
||
用法：
|
||
python3 scripts/fetch_images.py --query "Petronas Towers" "Langkawi beach" "George Town street" \
|
||
--out 项目/assets/img --count 2 --width 1600
|
||
|
||
每个 query 取前 count 张、缩放到 width、下载到 out，并打印清单（路径 | 许可 | 作者 | 来源页）便于诚实性核对。
|
||
全部抓不到 → 退出码 1，提示走 Phase 3.5 取图三级兜底（Unsplash/Pexels → 生图 → 诚实 placeholder）。
|
||
"""
|
||
import argparse, json, os, re, sys, urllib.parse, urllib.request
|
||
|
||
# ① 清代理：本机 curl/urllib 走代理会 TLS 炸（见 memory feedback_gemini_proxy）
|
||
for _k in ("ALL_PROXY", "all_proxy", "HTTP_PROXY", "http_proxy", "HTTPS_PROXY", "https_proxy"):
|
||
os.environ.pop(_k, None)
|
||
|
||
API = "https://commons.wikimedia.org/w/api.php"
|
||
# ② 合规 User-Agent 是硬性要求，否则 Wikimedia 返 429
|
||
UA = "huashu-design-image-fetcher/1.0 (https://huasheng.ai; skill contact)"
|
||
|
||
|
||
def _api_get(params):
|
||
url = API + "?" + urllib.parse.urlencode(params)
|
||
req = urllib.request.Request(url, headers={"User-Agent": UA})
|
||
with urllib.request.urlopen(req, timeout=30) as r:
|
||
return json.load(r)
|
||
|
||
|
||
def _safe(name):
|
||
return re.sub(r"[^\w\-.]", "_", name)[:60]
|
||
|
||
|
||
def fetch(query, out, count, width):
|
||
params = {
|
||
"action": "query", "format": "json", "generator": "search",
|
||
"gsrsearch": query, "gsrnamespace": 6, "gsrlimit": count,
|
||
"prop": "imageinfo", "iiprop": "url|extmetadata", "iiurlwidth": width,
|
||
}
|
||
try:
|
||
data = _api_get(params)
|
||
except Exception as e:
|
||
print(f"[FAIL search] {query}: {e}", file=sys.stderr)
|
||
return []
|
||
pages = (data.get("query", {}) or {}).get("pages", {})
|
||
got = []
|
||
for p in list(pages.values())[:count]:
|
||
ii = (p.get("imageinfo") or [{}])[0]
|
||
thumb = ii.get("thumburl") or ii.get("url")
|
||
if not thumb:
|
||
continue
|
||
meta = ii.get("extmetadata", {}) or {}
|
||
lic = (meta.get("LicenseShortName", {}) or {}).get("value", "?")
|
||
artist = re.sub("<[^>]+>", "", (meta.get("Artist", {}) or {}).get("value", "?")).strip()
|
||
ext = os.path.splitext(thumb)[1].split("?")[0] or ".jpg"
|
||
fn = _safe(query) + "_" + _safe(p.get("title", "img").replace("File:", ""))
|
||
fn = os.path.splitext(fn)[0][:55] + ext
|
||
path = os.path.join(out, fn)
|
||
try:
|
||
req = urllib.request.Request(thumb, headers={"User-Agent": UA})
|
||
with urllib.request.urlopen(req, timeout=60) as r, open(path, "wb") as f:
|
||
f.write(r.read())
|
||
got.append(path)
|
||
print(f"[OK] {path} | {lic} | {artist} | {ii.get('descriptionurl','')}")
|
||
except Exception as e:
|
||
print(f"[FAIL dl] {thumb}: {e}", file=sys.stderr)
|
||
if not got:
|
||
print(f"[EMPTY] 「{query}」没抓到——换关键词，或走 Phase 3.5 兜底", file=sys.stderr)
|
||
return got
|
||
|
||
|
||
def main():
|
||
ap = argparse.ArgumentParser(description="Wikimedia Commons 真图抓取（huashu-design Phase 3.5）")
|
||
ap.add_argument("--query", nargs="+", required=True, help="一个或多个英文关键词（英文命中率高）")
|
||
ap.add_argument("--out", required=True, help="输出目录（建议 项目/assets/img）")
|
||
ap.add_argument("--count", type=int, default=2, help="每个关键词抓几张（默认 2）")
|
||
ap.add_argument("--width", type=int, default=1600, help="缩放宽度 px（默认 1600）")
|
||
a = ap.parse_args()
|
||
os.makedirs(a.out, exist_ok=True)
|
||
allgot = []
|
||
for q in a.query:
|
||
allgot += fetch(q, a.out, a.count, a.width)
|
||
print(f"\n=== 共下载 {len(allgot)} 张到 {a.out} ===")
|
||
print("⚠️ 诚实性核对：去掉每张图信息是否有损？许可是否允许用途？不合适的删掉。")
|
||
if not allgot:
|
||
print("❌ 全部失败 → 走 Phase 3.5 取图三级兜底（Unsplash/Pexels → 生图 → 诚实 placeholder，不卡流程）", file=sys.stderr)
|
||
sys.exit(1)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|