Spaces:
Runtime error
Runtime error
File size: 12,074 Bytes
f166d1d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 | """
Step 1: 从视频目录中提取每个三级文件夹(采集时段)的代表帧
- 遍历所有 L1/L2/L3 结构,找最后一个 MP4,抽取第 0 帧
- 统计该 L3 下所有 MP4 的总时长(分钟)
- 输出图片到 /mnt/nas/Processing_data/Label_cites/ 保持相同目录结构
- 生成索引文件 sessions_index.json 供 step2 使用
- 每个 (L2, 日期) 组生成日代表图:取当天首/中/尾三个 L3 的图片
特殊结构处理:
- ShenZhen: L3 下还有 L4 子文件夹,递归找 MP4
- Haerbin: MP4 直接在 L2(无 L3),L2 本身作为唯一 session
- Hongkong: L2 下有 'video' 子文件夹作为 pseudo-L3
运行:
conda activate nds && python step1_extract_frames.py # 断点续跑
conda activate nds && python step1_extract_frames.py --reset # 清空重新提取
"""
import os
import json
import re
import sys
from collections import defaultdict
from pathlib import Path
from concurrent.futures import ProcessPoolExecutor, as_completed
import cv2
NUM_WORKERS = os.cpu_count() or 4
VIDEO_ROOT = "/mnt/nas/ADSafety/ADSafety/Aerial_Raw_Videos/2025-Changchun-AerialVideo-V1/"
OUTPUT_ROOT = "/mnt/nas/Processing_data/Label_cites"
INDEX_FILE = "/home/stu6/Projects/LabelWork/sessions_index.json"
VIDEO_EXTS = {".mp4", ".MP4"}
# ─────────────────────────── 工具函数 ────────────────────────────
def find_mp4s(folder: Path) -> list:
"""递归找 folder 下所有 MP4 文件,按名称排序"""
found = []
for root, _, files in os.walk(folder):
for f in sorted(files):
if Path(f).suffix in VIDEO_EXTS:
found.append(Path(root) / f)
return sorted(found)
def get_video_duration_minutes(video_path: Path) -> float:
"""用 OpenCV 获取视频时长(分钟)"""
try:
cap = cv2.VideoCapture(str(video_path))
fps = cap.get(cv2.CAP_PROP_FPS)
frame_count = cap.get(cv2.CAP_PROP_FRAME_COUNT)
cap.release()
if fps > 0 and frame_count > 0:
return frame_count / fps / 60.0
except Exception:
pass
# 回退:文件大小估算(1GB ≈ 1分钟)
try:
return video_path.stat().st_size / (1024 ** 3)
except Exception:
return 0.0
def extract_first_frame(video_path: Path, output_jpg: Path) -> bool:
"""用 OpenCV 抽取视频第 0 帧为 JPG"""
output_jpg.parent.mkdir(parents=True, exist_ok=True)
if output_jpg.exists():
return True
try:
cap = cv2.VideoCapture(str(video_path))
ret, frame = cap.read()
cap.release()
if ret:
cv2.imwrite(str(output_jpg), frame)
return output_jpg.exists()
except Exception as e:
print(f" [ERROR] OpenCV 失败 {video_path.name}: {e}")
return False
def parse_collection_time(video_name: str) -> str:
"""从文件名解析采集时间,格式:DJI_20250711155546_0001_V → 2025年7月11日 15:55"""
m = re.search(r"(\d{4})(\d{2})(\d{2})(\d{2})(\d{2})", video_name)
if m:
y, mo, d, h, mi = m.groups()
return f"{y}年{int(mo)}月{int(d)}日 {int(h):02d}:{int(mi):02d}"
return ""
def parse_date_str(name: str) -> str:
"""从名称中提取日期字符串 YYYYMMDD,用于按天分组"""
m = re.search(r"(\d{8})", name)
return m.group(1) if m else ""
# ──────────────────────────── 核心处理 ────────────────────────────
def process_session(l1_name: str, l2_name: str, l3_name: str,
session_folder: Path, relative_path: str):
"""
处理一个 session:找最后一个 MP4、提取其第 0 帧、算总时长
返回 session 信息字典,失败返回 None
"""
mp4s = find_mp4s(session_folder)
if not mp4s:
print(f" [SKIP] 无 MP4: {session_folder}")
return None
last_mp4 = mp4s[-1] # 取最后一个视频
video_name = last_mp4.name
output_jpg = Path(OUTPUT_ROOT) / relative_path / (last_mp4.stem + ".jpg")
ok = extract_first_frame(last_mp4, output_jpg)
if not ok:
print(f" [ERROR] 帧提取失败: {relative_path}")
return None
total_duration = sum(get_video_duration_minutes(mp4) for mp4 in mp4s)
collection_time = parse_collection_time(video_name)
return {
"l1": l1_name,
"l2": l2_name,
"l3": l3_name,
"folder_path": relative_path,
"video_name": video_name,
"location_name": l2_name,
"collection_time": collection_time,
"duration": round(total_duration, 4),
"image_path": str(output_jpg),
"mp4_count": len(mp4s),
}
def build_day_summaries(sessions: list) -> list:
"""
对每个 (l1, l2, 日期) 分组,选出首/中/尾三个 session 作为当天代表图。
日期从 l3 名称解析,回退到 video_name。
"""
groups = defaultdict(list)
for s in sessions:
date = parse_date_str(s["l3"]) or parse_date_str(s["video_name"])
groups[(s["l1"], s["l2"], date)].append(s)
summaries = []
for (l1, l2, date), group in sorted(groups.items()):
group = sorted(group, key=lambda s: s["l3"])
n = len(group)
# 选首/中/尾,去重
candidates = [group[0], group[n // 2], group[-1]]
seen, picks = set(), []
for p in candidates:
if p["folder_path"] not in seen:
seen.add(p["folder_path"])
picks.append(p)
date_display = ""
if len(date) == 8:
date_display = f"{date[:4]}年{int(date[4:6])}月{int(date[6:8])}日"
total_duration = round(sum(s["duration"] for s in group), 4)
representatives = [
{
"folder_path": p["folder_path"],
"image_path": p["image_path"],
"collection_time": p["collection_time"],
"duration": p["duration"],
}
for p in picks
]
summaries.append({
"l1": l1,
"l2": l2,
"date": date,
"date_display": date_display,
"session_count": n,
"total_duration": total_duration,
"representatives": representatives,
})
return summaries
# ────────────────────────── 目录扫描 ──────────────────────────────
def collect_tasks(video_root: Path, done_paths: set) -> list:
"""
Phase 1:单线程遍历目录,收集所有待处理任务。
返回 list of (l1_name, l2_name, l3_name, session_folder, rel)
"""
tasks = []
for l1_dir in sorted(video_root.iterdir()):
if not l1_dir.is_dir():
continue
l1_name = l1_dir.name
for l2_dir in sorted(l1_dir.iterdir()):
if not l2_dir.is_dir():
continue
l2_name = l2_dir.name
# ── 特殊情况1:Haerbin —— MP4 直接在 L2,无 L3 ──
mp4s_in_l2 = [f for f in l2_dir.iterdir()
if f.is_file() and f.suffix in VIDEO_EXTS]
if mp4s_in_l2:
rel = f"{l1_name}/{l2_name}"
if rel not in done_paths:
tasks.append((l1_name, l2_name, l2_name, l2_dir, rel))
continue
# ── 特殊情况2:Hongkong —— L2 下有 'video' 子文件夹 ──
video_subdir = l2_dir / "video"
if video_subdir.is_dir():
rel = f"{l1_name}/{l2_name}/video"
if rel not in done_paths:
tasks.append((l1_name, l2_name, "video", video_subdir, rel))
continue
# ── 标准情况:L3 子文件夹 ──
for l3_dir in sorted(d for d in l2_dir.iterdir() if d.is_dir()):
l3_name = l3_dir.name
rel = f"{l1_name}/{l2_name}/{l3_name}"
if rel not in done_paths:
tasks.append((l1_name, l2_name, l3_name, l3_dir, rel))
return tasks
def _worker(args):
"""进程池工作函数"""
l1_name, l2_name, l3_name, session_folder, rel = args
return process_session(l1_name, l2_name, l3_name, session_folder, rel)
# ──────────────────────────── 主流程 ──────────────────────────────
def main():
reset = "--reset" in sys.argv
video_root = Path(VIDEO_ROOT)
if reset:
print("[RESET] 删除已有图片和索引...")
if Path(OUTPUT_ROOT).exists():
jpgs = list(Path(OUTPUT_ROOT).rglob("*.jpg")) + list(Path(OUTPUT_ROOT).rglob("*.JPG"))
for jpg in jpgs:
jpg.unlink(missing_ok=True)
print(f" 已删除 {len(jpgs)} 张图片")
if Path(INDEX_FILE).exists():
Path(INDEX_FILE).unlink()
print(f" 已删除 {INDEX_FILE}")
# 加载已有索引(支持断点续跑)
sessions = []
done_paths = set()
if Path(INDEX_FILE).exists():
with open(INDEX_FILE, encoding="utf-8") as f:
existing = json.load(f)
sessions = existing.get("sessions", existing) if isinstance(existing, dict) else existing
done_paths = {s["folder_path"] for s in sessions}
print(f"[断点续跑] 已有 {len(done_paths)} 条记录")
# Phase 1:收集任务
print("Phase 1: 扫描目录结构...")
tasks = collect_tasks(video_root, done_paths)
print(f" 待处理 session 数: {len(tasks)},使用 {NUM_WORKERS} 个进程")
if tasks:
# Phase 2:并行提取帧
print("Phase 2: 并行提取帧...")
SAVE_INTERVAL = 20
completed = 0
with ProcessPoolExecutor(max_workers=NUM_WORKERS) as executor:
futures = {executor.submit(_worker, t): t[4] for t in tasks}
for future in as_completed(futures):
rel = futures[future]
try:
session = future.result()
except Exception as e:
print(f" [ERROR] {rel}: {e}")
session = None
if session:
sessions.append(session)
done_paths.add(rel)
print(f" [OK] {rel}")
completed += 1
if completed % SAVE_INTERVAL == 0:
_save_index(sessions)
print(f" [保存] {completed}/{len(tasks)} 完成")
# Phase 3:生成日代表图索引
print("Phase 3: 生成日代表图索引...")
day_summaries = build_day_summaries(sessions)
print(f" 共 {len(day_summaries)} 个 (场景, 日期) 组合")
_save_index(sessions, day_summaries)
# 统计汇报
print(f"\n✅ 完成: {len(sessions)} 个 sessions,{len(day_summaries)} 个日期组")
by_l1 = defaultdict(lambda: {"count": 0, "duration": 0.0})
for s in sessions:
by_l1[s["l1"]]["count"] += 1
by_l1[s["l1"]]["duration"] += s["duration"]
for k in sorted(by_l1):
print(f" {k}: {by_l1[k]['count']} sessions, {by_l1[k]['duration']:.1f} 分钟")
def _save_index(sessions: list, day_summaries: list = None):
data = {"sessions": sessions}
if day_summaries is not None:
data["day_summaries"] = day_summaries
elif Path(INDEX_FILE).exists():
# 保留已有的 day_summaries
with open(INDEX_FILE, encoding="utf-8") as f:
old = json.load(f)
if isinstance(old, dict) and "day_summaries" in old:
data["day_summaries"] = old["day_summaries"]
with open(INDEX_FILE, "w", encoding="utf-8") as f:
json.dump(data, f, ensure_ascii=False, indent=2)
if __name__ == "__main__":
main()
|