6.2
This commit is contained in:
hsz
@@ -0,0 +1,564 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
仅在「时间段 txt」内跑：人手检测 → **逐帧**好/坏门控（**top1 为 good 且 top1conf>阈值**，默认阈值 0.9）
|
||||
→ 仅通过的帧跑 41 类耗材分类；（可选）仅保留 **耗材 softmax 最大值 > --haocai-min-conf** 的帧；
|
||||
对保留帧的标签序列做 **滑动窗口多数票平滑**，再 **`consumable` 取平滑后序列众数**。
|
||||
|
||||
**avg_softmax_*** ：仅对上述「高置信耗材帧」统计；类别为 softmax 均值分布前三；置信度为三档边际 softmax 在时间上的平均。
|
||||
|
||||
不扫全片；每段从视频中按起止时间解码。
|
||||
|
||||
用法（建议在 yolo 环境）:
|
||||
python code/video_clip_cls/infer_single_0506/run_segments_consumable_vote.py \\
|
||||
--segments .../03视频_segments_mutual_exclusive_score_gt_0.1.txt \\
|
||||
--video .../03视频.mp4 \\
|
||||
--out .../03视频_segments_consumables.txt
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import sys
|
||||
from collections import Counter
|
||||
from pathlib import Path
|
||||
|
||||
import cv2
|
||||
import numpy as np
|
||||
from ultralytics import YOLO
|
||||
|
||||
for _repo in Path(__file__).resolve().parents:
|
||||
if (_repo / "repo_root.py").is_file() and (_repo / "dataset.py").is_file():
|
||||
if str(_repo) not in sys.path:
|
||||
sys.path.insert(0, str(_repo))
|
||||
break
|
||||
else:
|
||||
raise RuntimeError("未定位到仓库 code/ 根目录")
|
||||
|
||||
from repo_root import CODE_ROOT # noqa: E402
|
||||
|
||||
|
||||
def parse_segments_txt(path: Path) -> list[tuple[int, float, float]]:
|
||||
rows: list[tuple[int, float, float]] = []
|
||||
for raw in path.read_text(encoding="utf-8").splitlines():
|
||||
if not raw.strip() or raw.lower().startswith("rank"):
|
||||
continue
|
||||
parts = raw.split("\t")
|
||||
if len(parts) < 4:
|
||||
continue
|
||||
rank = int(parts[0].strip())
|
||||
t0 = float(parts[1].strip())
|
||||
t1 = float(parts[2].strip())
|
||||
rows.append((rank, t0, t1))
|
||||
return rows
|
||||
|
||||
|
||||
def collect_hand_boxes(det_model: YOLO, boxes) -> list[list[float]]:
|
||||
names = det_model.names
|
||||
out: list[list[float]] = []
|
||||
for box in boxes:
|
||||
cid = int(box.cls[0])
|
||||
label = names.get(cid, "")
|
||||
if label == "hand":
|
||||
out.append(box.xyxy[0].tolist())
|
||||
return out
|
||||
|
||||
|
||||
def pad_box(
|
||||
xyxy: list[float], img_w: int, img_h: int, pad_ratio: float
|
||||
) -> tuple[int, int, int, int]:
|
||||
"""四向等比外扩（legacy；Phase2 现用 pad_box_bottom_only）。"""
|
||||
x1, y1, x2, y2 = xyxy
|
||||
bw, bh = x2 - x1, y2 - y1
|
||||
px, py = bw * pad_ratio, bh * pad_ratio
|
||||
return (
|
||||
max(0, int(x1 - px)),
|
||||
max(0, int(y1 - py)),
|
||||
min(img_w, int(x2 + px)),
|
||||
min(img_h, int(y2 + py)),
|
||||
)
|
||||
|
||||
|
||||
def pad_box_bottom_only(
|
||||
xyxy: list[float], img_w: int, img_h: int, bottom_ratio: float
|
||||
) -> tuple[int, int, int, int]:
|
||||
"""紧框 union 后仅向下延伸：y2 += 框高 * bottom_ratio；上/左/右不变。"""
|
||||
x1, y1, x2, y2 = xyxy
|
||||
bh = y2 - y1
|
||||
dy = bh * float(bottom_ratio)
|
||||
return (
|
||||
max(0, int(x1)),
|
||||
max(0, int(y1)),
|
||||
min(img_w, int(x2)),
|
||||
min(img_h, int(y2 + dy)),
|
||||
)
|
||||
|
||||
|
||||
def largest_hand(hands: list[list[float]]) -> list[float]:
|
||||
def area(b: list[float]) -> float:
|
||||
return max(0.0, b[2] - b[0]) * max(0.0, b[3] - b[1])
|
||||
|
||||
return max(hands, key=area)
|
||||
|
||||
|
||||
def _float_top1conf(pr) -> float:
|
||||
tc = pr.top1conf
|
||||
if tc is None:
|
||||
return 0.0
|
||||
if isinstance(tc, (float, int, np.floating)):
|
||||
return float(tc)
|
||||
return float(tc.detach().float().cpu().item())
|
||||
|
||||
|
||||
def passes_good_gate_top1_conf(
|
||||
gb_model: YOLO,
|
||||
crop: np.ndarray,
|
||||
gb_names: dict,
|
||||
imgsz: int,
|
||||
top1_conf_must_exceed: float,
|
||||
) -> bool:
|
||||
"""好/坏分类：predicted top1 为 good，且 top1conf 严格大于给定阈值。"""
|
||||
if crop.size == 0:
|
||||
return False
|
||||
r = gb_model.predict(crop, imgsz=imgsz, verbose=False)[0]
|
||||
pr = r.probs
|
||||
if pr is None:
|
||||
return False
|
||||
tid = int(pr.top1)
|
||||
label = str(gb_names.get(tid, "")).strip().lower()
|
||||
conf = _float_top1conf(pr)
|
||||
return label == "good" and conf > top1_conf_must_exceed
|
||||
|
||||
|
||||
def haocai_softmax_probs(
|
||||
cls_model: YOLO, crop: np.ndarray, imgsz: int, n_cls: int
|
||||
) -> np.ndarray | None:
|
||||
"""耗材分类：返回长度 n_cls 的 softmax 概率向量（与模型 top1 一致）。"""
|
||||
if crop.size == 0:
|
||||
return None
|
||||
r = cls_model.predict(crop, imgsz=imgsz, verbose=False)[0]
|
||||
pr = r.probs
|
||||
if pr is None or pr.data is None:
|
||||
return None
|
||||
v = pr.data.detach().float().cpu().numpy().astype(np.float64).ravel()
|
||||
if v.size < n_cls:
|
||||
v = np.resize(v, n_cls)
|
||||
v = v[:n_cls].copy()
|
||||
s = float(np.sum(v))
|
||||
if s <= 1e-12:
|
||||
return None
|
||||
# 若未归一化则 softmax
|
||||
if abs(s - 1.0) > 0.08:
|
||||
v = v - float(np.max(v))
|
||||
e = np.exp(np.clip(v, -40.0, 40.0))
|
||||
out = e / float(np.sum(e))
|
||||
return out
|
||||
return v / s
|
||||
|
||||
|
||||
def _cls_name(names: dict, idx: int) -> str:
|
||||
return str(names.get(int(idx), str(idx)))
|
||||
|
||||
|
||||
def mean_softmax_top3(
|
||||
probs_list: list[np.ndarray], cls_names: dict
|
||||
) -> tuple[list[str], list[float]]:
|
||||
"""
|
||||
类名：多帧 softmax 按类逐维算术平均，在平均向量上取概率最大的前三类。
|
||||
|
||||
置信度（与类名解耦）：逐帧对 softmax 从高到低排序，取第 1/2/3 大的概率，
|
||||
再在各帧上对这三档分别做算术平均（「帧内边际 topk」的时间平均）。
|
||||
返回三个槽位（不足则用空字符串与 0.0 补齐）。
|
||||
"""
|
||||
names_out: list[str] = []
|
||||
probs_out: list[float] = []
|
||||
if not probs_list:
|
||||
for _ in range(3):
|
||||
names_out.append("")
|
||||
probs_out.append(0.0)
|
||||
return names_out, probs_out
|
||||
stacked = np.stack(probs_list, axis=0)
|
||||
p = np.mean(stacked, axis=0, dtype=np.float64)
|
||||
order = np.argsort(-p)
|
||||
for k in range(3):
|
||||
if k < order.size:
|
||||
j = int(order[k])
|
||||
names_out.append(_cls_name(cls_names, j))
|
||||
else:
|
||||
names_out.append("")
|
||||
# 逐帧降序 softmax，对第 1/2/3 档做时间平均
|
||||
row_sorted = np.sort(stacked, axis=1)[:, ::-1]
|
||||
n_cls = row_sorted.shape[1]
|
||||
for k in range(3):
|
||||
if k < n_cls:
|
||||
probs_out.append(float(np.mean(row_sorted[:, k], dtype=np.float64)))
|
||||
else:
|
||||
probs_out.append(0.0)
|
||||
return names_out, probs_out
|
||||
|
||||
|
||||
def smooth_labels_majority(labels: list[str], window: int) -> list[str]:
|
||||
"""
|
||||
对时间有序的类别名做平滑：对每个位置取以该位置为中心、长度为奇数 window 的邻域，
|
||||
用邻域内众数替换（打破平局时用最邻域计数最高者）。
|
||||
window<=1 时原样返回。
|
||||
"""
|
||||
if window <= 1 or not labels:
|
||||
return list(labels)
|
||||
w = window if window % 2 == 1 else window + 1
|
||||
half = w // 2
|
||||
n = len(labels)
|
||||
out: list[str] = []
|
||||
for i in range(n):
|
||||
lo = max(0, i - half)
|
||||
hi = min(n, i + half + 1)
|
||||
chunk = labels[lo:hi]
|
||||
top, _c = Counter(chunk).most_common(1)[0]
|
||||
out.append(top)
|
||||
return out
|
||||
|
||||
|
||||
def process_segment(
|
||||
cap: cv2.VideoCapture,
|
||||
det: YOLO,
|
||||
gb: YOLO,
|
||||
cls_m: YOLO,
|
||||
*,
|
||||
start_sec: float,
|
||||
end_sec: float,
|
||||
seek_margin_sec: float,
|
||||
det_conf: float,
|
||||
pad_ratio: float,
|
||||
imgsz_det: int,
|
||||
imgsz_cls: int,
|
||||
frame_stride: int,
|
||||
good_top1_conf_threshold: float,
|
||||
haocai_min_conf: float,
|
||||
smooth_label_window: int,
|
||||
gb_names: dict,
|
||||
cls_names: dict,
|
||||
) -> dict:
|
||||
# HEVC/部分 mp4：直接 seek 到 start 易产生坏参考帧；先往回跳再顺序解码丢到起点。
|
||||
probe_from = float(max(0.0, start_sec - seek_margin_sec))
|
||||
cap.set(cv2.CAP_PROP_POS_MSEC, probe_from * 1000.0)
|
||||
synced_frame: np.ndarray | None = None
|
||||
synced_t: float | None = None
|
||||
tol = 0.04
|
||||
while True:
|
||||
ok0, grab = cap.read()
|
||||
if not ok0 or grab is None:
|
||||
synced_frame, synced_t = None, None
|
||||
break
|
||||
t0 = float(cap.get(cv2.CAP_PROP_POS_MSEC)) / 1000.0
|
||||
if t0 + tol >= start_sec:
|
||||
synced_frame, synced_t = grab, t0
|
||||
break
|
||||
|
||||
w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
|
||||
h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
||||
|
||||
n_cls_key_max = max(int(k) for k in cls_names.keys())
|
||||
n_cls = n_cls_key_max + 1
|
||||
|
||||
n_hand_frames = 0
|
||||
# top1==good 且 top1conf>阈值的帧数（门控通过即计数，与是否成功得到 softmax 无关）
|
||||
n_gate_pass = 0
|
||||
cls_labels: list[str] = []
|
||||
cls_prob_rows: list[np.ndarray] = []
|
||||
frames_read_in_segment = 0
|
||||
|
||||
def one_frame(fr: np.ndarray, _t_abs: float) -> None:
|
||||
nonlocal frames_read_in_segment, n_hand_frames, n_gate_pass, cls_labels, cls_prob_rows
|
||||
frames_read_in_segment += 1
|
||||
if frame_stride > 1 and (frames_read_in_segment - 1) % frame_stride != 0:
|
||||
return
|
||||
|
||||
r0 = det.predict(
|
||||
fr,
|
||||
conf=det_conf,
|
||||
imgsz=imgsz_det,
|
||||
verbose=False,
|
||||
)[0]
|
||||
hands = collect_hand_boxes(det, r0.boxes) if r0.boxes else []
|
||||
if not hands:
|
||||
return
|
||||
|
||||
n_hand_frames += 1
|
||||
xyxy = largest_hand(hands)
|
||||
x1, y1, x2, y2 = pad_box(xyxy, w, h, pad_ratio)
|
||||
crop = fr[y1:y2, x1:x2]
|
||||
ok_gate = passes_good_gate_top1_conf(
|
||||
gb, crop, gb_names, imgsz_cls, good_top1_conf_threshold
|
||||
)
|
||||
if ok_gate:
|
||||
n_gate_pass += 1
|
||||
vec = haocai_softmax_probs(cls_m, crop, imgsz_cls, n_cls)
|
||||
if vec is not None:
|
||||
top_prob = float(np.max(vec))
|
||||
if top_prob <= haocai_min_conf:
|
||||
return
|
||||
cls_prob_rows.append(vec)
|
||||
cls_labels.append(_cls_name(cls_names, int(np.argmax(vec))))
|
||||
|
||||
if synced_frame is not None and synced_t is not None:
|
||||
if synced_t <= end_sec + 0.08:
|
||||
one_frame(synced_frame, synced_t)
|
||||
|
||||
while True:
|
||||
ok, frame = cap.read()
|
||||
if not ok or frame is None:
|
||||
break
|
||||
t = float(cap.get(cv2.CAP_PROP_POS_MSEC)) / 1000.0
|
||||
if t > end_sec + 0.08:
|
||||
break
|
||||
if t + 1e-6 < start_sec:
|
||||
continue
|
||||
one_frame(frame, t)
|
||||
|
||||
if n_hand_frames == 0:
|
||||
return {
|
||||
"consumable": "（段内未检测到手部）",
|
||||
"n_hand_frames": 0,
|
||||
"n_gate_pass": 0,
|
||||
"n_predictions": 0,
|
||||
"top_vote_count": 0,
|
||||
"avg_top1_cls": "",
|
||||
"avg_top1_prob": "",
|
||||
"avg_top2_cls": "",
|
||||
"avg_top2_prob": "",
|
||||
"avg_top3_cls": "",
|
||||
"avg_top3_prob": "",
|
||||
}
|
||||
|
||||
if not cls_labels:
|
||||
return {
|
||||
"consumable": (
|
||||
"（无满足条件的耗材帧：好帧置信度或未过门控"
|
||||
+ (
|
||||
"" if haocai_min_conf <= 0.0
|
||||
else "，或耗材 top1 softmax 不大于阈值"
|
||||
)
|
||||
+ "）"
|
||||
),
|
||||
"n_hand_frames": n_hand_frames,
|
||||
"n_gate_pass": n_gate_pass,
|
||||
"n_predictions": 0,
|
||||
"top_vote_count": 0,
|
||||
"avg_top1_cls": "",
|
||||
"avg_top1_prob": "",
|
||||
"avg_top2_cls": "",
|
||||
"avg_top2_prob": "",
|
||||
"avg_top3_cls": "",
|
||||
"avg_top3_prob": "",
|
||||
}
|
||||
|
||||
smoothed = smooth_labels_majority(cls_labels, smooth_label_window)
|
||||
top_name, vote_n = Counter(smoothed).most_common(1)[0]
|
||||
a1, ap1 = mean_softmax_top3(cls_prob_rows, cls_names)
|
||||
return {
|
||||
"consumable": top_name,
|
||||
"n_hand_frames": n_hand_frames,
|
||||
"n_gate_pass": n_gate_pass,
|
||||
"n_predictions": len(cls_labels),
|
||||
"top_vote_count": int(vote_n),
|
||||
"avg_top1_cls": a1[0],
|
||||
"avg_top1_prob": f"{ap1[0]:.6f}",
|
||||
"avg_top2_cls": a1[1],
|
||||
"avg_top2_prob": f"{ap1[1]:.6f}",
|
||||
"avg_top3_cls": a1[2],
|
||||
"avg_top3_prob": f"{ap1[2]:.6f}",
|
||||
}
|
||||
|
||||
|
||||
def main() -> int:
|
||||
ap = argparse.ArgumentParser(
|
||||
description="手检 + 逐帧 top1=good 且 top1conf>阈值门控 + 耗材分类；段内众数"
|
||||
)
|
||||
ap.add_argument(
|
||||
"--segments",
|
||||
type=Path,
|
||||
default=Path(__file__).resolve().parent
|
||||
/ "results"
|
||||
/ "03视频_segments_mutual_exclusive_score_gt_0.1.txt",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--video",
|
||||
type=Path,
|
||||
default=CODE_ROOT.parent
|
||||
/ "data/haocai/5月6号视频/5月6日第二次视频/03视频.mp4",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--hand-model",
|
||||
type=Path,
|
||||
default=CODE_ROOT
|
||||
/ "hand_detection/runs/hand_det_y11s_multiframe-better/weights/best.pt",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--goodbad-model",
|
||||
type=Path,
|
||||
default=CODE_ROOT
|
||||
/ "goodORbad_frame/runs/goodbad_frame_y11m_e50/weights/best.pt",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--haocai-model",
|
||||
type=Path,
|
||||
default=CODE_ROOT
|
||||
/ "haocai_classify/runs/haocai_cls_41cls_goodframe_lastest-0.95"
|
||||
/ "weights/best.pt",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--out",
|
||||
type=Path,
|
||||
default=Path(__file__).resolve().parent
|
||||
/ "results"
|
||||
/ "03视频_segments_consumables.txt",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--good-top1-conf-threshold",
|
||||
type=float,
|
||||
default=0.90,
|
||||
dest="good_top1_conf_threshold",
|
||||
help="逐帧：仅当 top1 为 good 且 top1conf **严格大于**该值时才跑耗材分类（默认对应 top1conf>0.9）",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--haocai-min-conf",
|
||||
type=float,
|
||||
default=0.0,
|
||||
metavar="P",
|
||||
help="耗材：仅 softmax 最大值 **严格大于** P 的帧计入标签与 softmax 统计（0 表示不按耗材置信度筛）",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--smooth-label-window",
|
||||
type=int,
|
||||
default=1,
|
||||
metavar="W",
|
||||
help="耗材标签平滑：长度为 W 的奇数滑动窗口内多数票（W≤1 不平滑）；众数取平滑后的序列",
|
||||
)
|
||||
ap.add_argument("--det-conf", type=float, default=0.5)
|
||||
ap.add_argument("--pad-ratio", type=float, default=0.30)
|
||||
ap.add_argument("--imgsz-det", type=int, default=640)
|
||||
ap.add_argument("--imgsz-cls", type=int, default=224)
|
||||
ap.add_argument(
|
||||
"--frame-stride",
|
||||
type=int,
|
||||
default=1,
|
||||
help=">1 时代码逐帧解码但每 N 帧推理一次（省算力，结论可能略粗糙）",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--seek-margin-sec",
|
||||
type=float,
|
||||
default=3.0,
|
||||
help="HEVC 等非关键帧 seek 时往回多跳若干秒再解码到段起点，减轻花屏",
|
||||
)
|
||||
args = ap.parse_args()
|
||||
|
||||
seg_path = args.segments.resolve()
|
||||
vid_path = args.video.resolve()
|
||||
if not seg_path.is_file():
|
||||
print("找不到时间段文件:", seg_path, file=sys.stderr)
|
||||
return 1
|
||||
if not vid_path.is_file():
|
||||
print("找不到视频:", vid_path, file=sys.stderr)
|
||||
return 1
|
||||
for pt, lab in (
|
||||
(args.hand_model, "hand"),
|
||||
(args.goodbad_model, "good/bad"),
|
||||
(args.haocai_model, "haocai cls"),
|
||||
):
|
||||
if not Path(pt).is_file():
|
||||
print(f"缺少{lab} 权重:", pt, file=sys.stderr)
|
||||
return 1
|
||||
|
||||
segments = parse_segments_txt(seg_path)
|
||||
if not segments:
|
||||
print("时间段为空:", seg_path, file=sys.stderr)
|
||||
return 1
|
||||
|
||||
print("加载模型…", flush=True)
|
||||
det = YOLO(str(args.hand_model))
|
||||
gb = YOLO(str(args.goodbad_model))
|
||||
cls_m = YOLO(str(args.haocai_model))
|
||||
gb_names = gb.names
|
||||
cls_names = cls_m.names
|
||||
|
||||
cap = cv2.VideoCapture(str(vid_path))
|
||||
if not cap.isOpened():
|
||||
print("无法打开视频:", vid_path, file=sys.stderr)
|
||||
return 1
|
||||
|
||||
sep = "\t"
|
||||
out_lines = [
|
||||
sep.join([
|
||||
"rank",
|
||||
"start_sec",
|
||||
"end_sec",
|
||||
"consumable",
|
||||
"n_hand_frames",
|
||||
"n_frames_top1_good_conf_gt_thresh",
|
||||
"n_consumable_predictions",
|
||||
"top_label_vote_count",
|
||||
"avg_softmax_top1_cls",
|
||||
"avg_softmax_top1_prob",
|
||||
"avg_softmax_top2_cls",
|
||||
"avg_softmax_top2_prob",
|
||||
"avg_softmax_top3_cls",
|
||||
"avg_softmax_top3_prob",
|
||||
])
|
||||
]
|
||||
|
||||
try:
|
||||
for rank, t0, t1 in segments:
|
||||
print(f"段落 rank={rank} [{t0:.3f},{t1:.3f}]s …", flush=True)
|
||||
info = process_segment(
|
||||
cap,
|
||||
det,
|
||||
gb,
|
||||
cls_m,
|
||||
start_sec=t0,
|
||||
end_sec=t1,
|
||||
seek_margin_sec=args.seek_margin_sec,
|
||||
det_conf=args.det_conf,
|
||||
pad_ratio=args.pad_ratio,
|
||||
imgsz_det=args.imgsz_det,
|
||||
imgsz_cls=args.imgsz_cls,
|
||||
frame_stride=max(1, args.frame_stride),
|
||||
good_top1_conf_threshold=args.good_top1_conf_threshold,
|
||||
haocai_min_conf=args.haocai_min_conf,
|
||||
smooth_label_window=max(1, args.smooth_label_window),
|
||||
gb_names=gb_names,
|
||||
cls_names=cls_names,
|
||||
)
|
||||
row = sep.join([
|
||||
str(rank),
|
||||
f"{t0:.6f}",
|
||||
f"{t1:.6f}",
|
||||
str(info["consumable"]),
|
||||
str(info["n_hand_frames"]),
|
||||
str(info["n_gate_pass"]),
|
||||
str(info["n_predictions"]),
|
||||
str(info["top_vote_count"]),
|
||||
info["avg_top1_cls"],
|
||||
info["avg_top1_prob"],
|
||||
info["avg_top2_cls"],
|
||||
info["avg_top2_prob"],
|
||||
info["avg_top3_cls"],
|
||||
info["avg_top3_prob"],
|
||||
])
|
||||
out_lines.append(row)
|
||||
print(
|
||||
f" -> {info['consumable']} "
|
||||
f"(votes {info['top_vote_count']}/{info['n_predictions']}, "
|
||||
f"goodgate {info['n_gate_pass']}/{info['n_hand_frames']} hand frames)",
|
||||
flush=True,
|
||||
)
|
||||
finally:
|
||||
cap.release()
|
||||
|
||||
out_path = args.out.resolve()
|
||||
out_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
out_path.write_text("\n".join(out_lines) + "\n", encoding="utf-8")
|
||||
print("已写出:", out_path, flush=True)
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Reference in New Issue
Block a user