463 lines
18 KiB
Python
463 lines
18 KiB
Python
from lib.json_fun import f_detections_to_objects, load_json_data
|
||
from lib.qwen_fun import get_annnotated_frame_for_ai_without_xyxy, hazard_inspection, merge_conflict_inspection_data, report_generator, search_knowledge_base
|
||
from encodings.punycode import T
|
||
|
||
from lib.qwen_fun_vid import generate_video_to_objects
|
||
"""
|
||
测试 在给定标注框与类别，原始视频经过转换之后，AI能否准确识别物体特征
|
||
"""
|
||
|
||
from tkinter import N
|
||
from datetime import datetime
|
||
import json
|
||
import os
|
||
import re
|
||
import sys
|
||
import subprocess
|
||
import importlib.util
|
||
from typing import Any
|
||
from lib.qwen_fun import save_json_to_file, upload_files_and_get_urls_concurrently
|
||
from lib.sam3 import SAM3
|
||
import cv2
|
||
import json
|
||
import numpy as np
|
||
from pathlib import Path
|
||
import gradio as gr
|
||
|
||
fps = 0
|
||
VIDEO_FOLDER: str = "input"
|
||
|
||
def load_file_list() -> list[str]:
|
||
"""
|
||
加载指定文件夹下的所有文件名称（不含子目录）
|
||
返回：文件名列表
|
||
"""
|
||
file_names: list[str] = []
|
||
|
||
if not os.path.isdir(VIDEO_FOLDER):
|
||
return file_names
|
||
|
||
for item in os.listdir(VIDEO_FOLDER):
|
||
item_path: str = os.path.join(VIDEO_FOLDER, item)
|
||
if os.path.isfile(item_path):
|
||
file_names.append(item)
|
||
|
||
return file_names
|
||
|
||
def reload_files():
|
||
"""
|
||
刷新文件列表并设置默认值
|
||
"""
|
||
file_list = load_file_list()
|
||
default_value = file_list[0] if file_list else None
|
||
return gr.update(choices=file_list, value=default_value)
|
||
|
||
def get_full_vid_path(vid_file: str) -> str:
|
||
"""
|
||
获取视频绝对路径
|
||
"""
|
||
full_path: str = os.path.join(os.getcwd(), VIDEO_FOLDER, vid_file)
|
||
gr.Info(full_path)
|
||
return full_path
|
||
|
||
def update_preview(frame_idx: int, vid_file: str):
|
||
vid_name: str = Path(vid_file).stem
|
||
output_dir: str = f"output/{vid_name}"
|
||
global fps
|
||
idx = int(frame_idx // fps)
|
||
img_path = f"{output_dir}/boxes/frame_{idx:04d}.jpg"
|
||
|
||
with open(f"{output_dir}/hazard_inspection.json", "r", encoding="utf-8") as f:
|
||
result = f.read()
|
||
|
||
dic = json.loads(result)
|
||
class_tag = get_class_tag_by_frame(dic, frame_idx, fps)
|
||
|
||
return img_path, class_tag
|
||
|
||
def run(vid_file: str, run_sam3: bool = True, run_inspection: bool = True, gen_report: bool = True):
|
||
"""
|
||
把 batch_run_videos_then_analyze_images.py 的全部功能整合到这一个函数里。
|
||
针对 Gradio 选中的单个视频，执行：1.py 生成 track 图片 -> analyze_track_images_with_llm.py 检测图片。
|
||
|
||
参数（保持与原 run 一致，方便 Gradio 直接复用）：
|
||
vid_file: input/ 目录下的视频文件名
|
||
run_sam3: 是否运行 1.py 生成 track 图片目录（沿用原变量名做开关）
|
||
run_inspection: 是否运行 analyze 图片检测脚本
|
||
gen_report: 占位；analyze 脚本会自行生成 Word/JSON 报告
|
||
"""
|
||
|
||
# ============================================================================
|
||
# 【配置区域】（来自 batch_run_videos_then_analyze_images.py）
|
||
# ============================================================================
|
||
VIDEO_DIR = r""
|
||
VIDEO_PATHS: list[str] = [r"C:\factory-inspection\videos\MOV00001_20260213_103217_fixed.mp4"]
|
||
YOLO_MODEL_PATH = r"C:\factory-inspection\yolo\best.pt"
|
||
TRACK_SCRIPT_PATH = r"C:\factory-inspection\1.py"
|
||
ANALYZE_SCRIPT_PATH = r"C:\factory-inspection\scripts\analyze_track_images_with_llm.py"
|
||
OUTPUT_ROOT = r"C:\factory-inspection\batch_video_image_reports"
|
||
VIDEO_EXTS = {".mp4", ".avi", ".mov", ".mkv", ".MP4", ".AVI", ".MOV", ".MKV"}
|
||
# ============================================================================
|
||
|
||
def _safe_name(path: Path) -> str:
|
||
return re.sub(r"[^0-9A-Za-z\u4e00-\u9fff_.-]+", "_", path.stem or "video")
|
||
|
||
def _load_module(script_path: Path) -> Any:
|
||
spec = importlib.util.spec_from_file_location("factory_video_tracker", script_path)
|
||
if spec is None or spec.loader is None:
|
||
raise RuntimeError(f"无法加载脚本：{script_path}")
|
||
module = importlib.util.module_from_spec(spec)
|
||
sys.modules["factory_video_tracker"] = module
|
||
spec.loader.exec_module(module)
|
||
return module
|
||
|
||
def _collect_videos(video_dir: str, video_paths: list[str]) -> list[Path]:
|
||
videos: list[Path] = []
|
||
for raw in video_paths:
|
||
path = Path(raw)
|
||
if path.is_file():
|
||
videos.append(path)
|
||
|
||
if video_dir:
|
||
root = Path(video_dir)
|
||
if root.is_dir():
|
||
for path in sorted(root.rglob("*")):
|
||
if path.suffix in VIDEO_EXTS and path.is_file():
|
||
videos.append(path)
|
||
|
||
# 去重但保持顺序
|
||
result: list[Path] = []
|
||
seen: set[str] = set()
|
||
for video in videos:
|
||
key = str(video.resolve()).lower()
|
||
if key not in seen:
|
||
seen.add(key)
|
||
result.append(video)
|
||
return result
|
||
|
||
def _parse_class_id_and_name(value: Any) -> tuple[int | None, str]:
|
||
text = str(value or "").strip()
|
||
match = re.match(r"^\s*(\d+)\s+(.+?)\s*$", text)
|
||
if match:
|
||
return int(match.group(1)), match.group(2).strip()
|
||
match = re.match(r"^\s*(\d+)\s*$", text)
|
||
if match:
|
||
class_id = int(match.group(1))
|
||
return class_id, f"class_{class_id}"
|
||
return None, text or "未知目标"
|
||
|
||
def _parse_bbox(value: Any) -> list[float] | None:
|
||
if isinstance(value, list) and len(value) == 4:
|
||
return [float(x) for x in value]
|
||
if isinstance(value, str):
|
||
parts = [x.strip() for x in value.split(",") if x.strip()]
|
||
if len(parts) == 4:
|
||
return [float(x) for x in parts]
|
||
return None
|
||
|
||
def _track_id_from_path(json_path: Path) -> int | None:
|
||
match = re.search(r"track_(\d+)", json_path.stem) or re.search(r"track_(\d+)", json_path.parent.name)
|
||
return int(match.group(1)) if match else None
|
||
|
||
def _normalize_1py_jsons(track_dir: Path) -> None:
|
||
"""1.py 可能输出单个 dict；analyze 脚本常用的是 list。这里做格式转换。"""
|
||
for json_path in track_dir.rglob("*.json"):
|
||
try:
|
||
data = json.loads(json_path.read_text(encoding="utf-8"))
|
||
except Exception as exc:
|
||
print(f"跳过异常 JSON：{json_path}，原因：{exc}")
|
||
continue
|
||
if isinstance(data, list):
|
||
continue
|
||
if not isinstance(data, dict):
|
||
continue
|
||
|
||
class_id, class_name = _parse_class_id_and_name(data.get("class_id") or data.get("class_name"))
|
||
normalized = [{
|
||
"xyxy": _parse_bbox(data.get("xyxy") or data.get("bbox") or data.get("location")),
|
||
"confidence": data.get("confidence", data.get("conf", 0.0)),
|
||
"track_id": data.get("track_id", _track_id_from_path(json_path)),
|
||
"class_id": class_id,
|
||
"class_str": class_name,
|
||
"start_frame": data.get("start_frame"),
|
||
"end_frame": data.get("end_frame"),
|
||
"start_sec": data.get("start_sec"),
|
||
}]
|
||
json_path.write_text(json.dumps(normalized, ensure_ascii=False, indent=2), encoding="utf-8")
|
||
|
||
def _video_to_tracks(video_path: Path, yolo_model_path: Path, output_dir: Path) -> Path:
|
||
module = _load_module(Path(TRACK_SCRIPT_PATH))
|
||
config = module.InferenceConfig()
|
||
|
||
# 覆盖 1.py 里的硬编码路径
|
||
config.VIDEO_PATH = str(video_path)
|
||
config.MODEL_PATH = str(yolo_model_path)
|
||
|
||
tracks_dir = output_dir / "tracks"
|
||
config.BEST_FRAME_DIR = str(tracks_dir)
|
||
config.OUTPUT_DIR = str(output_dir / "tracker_output")
|
||
|
||
tracker = module.VideoTracker(config, output_path=str(output_dir / "tracked_video.mp4"))
|
||
tracker.track_video()
|
||
_normalize_1py_jsons(tracks_dir)
|
||
return tracks_dir
|
||
|
||
def _run_image_analyze(track_dir: Path, report_dir: Path) -> None:
|
||
cmd = [
|
||
sys.executable,
|
||
str(Path(ANALYZE_SCRIPT_PATH)),
|
||
"--input",
|
||
str(track_dir),
|
||
"--output",
|
||
str(report_dir),
|
||
]
|
||
env = os.environ.copy()
|
||
env.setdefault("PYTHONUTF8", "1")
|
||
env.setdefault("PYTHONIOENCODING", "utf-8")
|
||
# 防止子进程出现 is_available=True 但 device_count=0 的诡异情况
|
||
env.pop("CUDA_VISIBLE_DEVICES", None)
|
||
print("执行图片检测：", " ".join(f'"{x}"' if " " in x else x for x in cmd))
|
||
code = subprocess.call(cmd, env=env)
|
||
if code != 0:
|
||
raise RuntimeError(f"图片检测失败，退出码：{code}")
|
||
|
||
def _process_one_video(video_path: Path, yolo_model_path: Path, output_root: Path) -> tuple[Path, Path]:
|
||
run_name = f"{_safe_name(video_path)}_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
|
||
run_dir = output_root / run_name
|
||
report_dir = run_dir / "report"
|
||
run_dir.mkdir(parents=True, exist_ok=True)
|
||
report_dir.mkdir(parents=True, exist_ok=True)
|
||
|
||
print("=" * 60)
|
||
print(f"处理视频：{video_path}")
|
||
|
||
if run_sam3:
|
||
print("第 1 步：调用 1.py 生成 track 图片目录")
|
||
tracks_dir = _video_to_tracks(video_path, yolo_model_path, run_dir)
|
||
print(f"track 图片目录：{tracks_dir}")
|
||
else:
|
||
tracks_dir = run_dir / "tracks"
|
||
print(f"跳过 1.py，使用已存在的 track 图片目录：{tracks_dir}")
|
||
|
||
if run_inspection or gen_report:
|
||
print("第 2 步：调用 analyze_track_images_with_llm.py 检测图片")
|
||
_run_image_analyze(tracks_dir, report_dir)
|
||
print(f"报告目录：{report_dir}")
|
||
else:
|
||
print("跳过图片检测")
|
||
|
||
print("=" * 60)
|
||
return tracks_dir, report_dir
|
||
|
||
# ===== 主体逻辑：单视频走单视频分支，多视频/批量收集逻辑也保留可用 =====
|
||
selected_video: Path = Path(VIDEO_FOLDER) / vid_file if vid_file else Path()
|
||
|
||
if not selected_video.is_file():
|
||
# Gradio 没选到时，退回 batch 脚本配置的 VIDEO_DIR / VIDEO_PATHS
|
||
videos = _collect_videos(VIDEO_DIR, VIDEO_PATHS)
|
||
if not videos:
|
||
raise gr.Error(
|
||
f"找不到视频：{selected_video}；且 VIDEO_DIR / VIDEO_PATHS 都为空。"
|
||
)
|
||
selected_video = videos[0]
|
||
|
||
yolo_model = Path(str(YOLO_MODEL_PATH).strip().strip('"'))
|
||
if not yolo_model.is_file():
|
||
raise gr.Error(f"YOLO 权重不存在：{yolo_model}，请设置 YOLO_MODEL_PATH。")
|
||
|
||
output_root = Path(OUTPUT_ROOT)
|
||
output_root.mkdir(parents=True, exist_ok=True)
|
||
|
||
tracks_dir, report_dir = _process_one_video(selected_video, yolo_model, output_root)
|
||
print(f"视频处理完成：{selected_video}")
|
||
|
||
# ===== 给 Gradio 返回 4 元组（沿用原 run 的输出结构） =====
|
||
# 取一张 track 图片作为预览
|
||
preview_path: str = ""
|
||
for img in tracks_dir.rglob("*.jpg"):
|
||
preview_path = str(img)
|
||
break
|
||
if not preview_path:
|
||
for img in tracks_dir.rglob("*.png"):
|
||
preview_path = str(img)
|
||
break
|
||
|
||
# 尝试读取 analyze 输出的 JSON 报告
|
||
json_result: dict | list = {}
|
||
for jp in sorted(report_dir.rglob("*.json")):
|
||
try:
|
||
json_result = json.loads(Path(jp).read_text(encoding="utf-8"))
|
||
break
|
||
except Exception:
|
||
continue
|
||
|
||
status_text = (
|
||
f"处理完成。\n"
|
||
f"视频：{selected_video}\n"
|
||
f"Track 目录：{tracks_dir}\n"
|
||
f"报告目录：{report_dir}"
|
||
)
|
||
|
||
return gr.update(maximum=0, value=0, step=1), preview_path, status_text, json_result
|
||
|
||
|
||
|
||
def get_vid_dict(vid_dir: str, obj_dict: dict, video_url_file: str, use_url_cache: bool) -> dict:
|
||
"""获取视频字典
|
||
参数:
|
||
vid_dir: 物体视频存放目录
|
||
obj_dict: 物品字典，键为物品TrackID，值为物品信息
|
||
video_url_file: 视频URL文件路径
|
||
use_url_cache: 是否使用缓存的URL
|
||
返回:
|
||
视频字典，键为物品TrackID，值为视频本地地址和视频URL
|
||
"""
|
||
|
||
vid_dict = {}
|
||
|
||
# 检查URL是否已存在
|
||
if use_url_cache:
|
||
if os.path.exists(video_url_file):
|
||
try:
|
||
with open(video_url_file, "r", encoding="utf-8") as f:
|
||
loaded_data = json.load(f)
|
||
# 验证数据结构：确保值是字典且包含必要键
|
||
for track_id, value in loaded_data.items():
|
||
if isinstance(value, dict) and "vid_path" in value and "vid_url" in value:
|
||
vid_dict[track_id] = value
|
||
else:
|
||
print(f"跳过无效的视频信息: {track_id} -> {value}")
|
||
except Exception as e:
|
||
print(f"读取URL文件失败: {e}")
|
||
|
||
# 初始化未存在的track_id
|
||
for track_id in obj_dict.keys():
|
||
if track_id in vid_dict:
|
||
continue
|
||
try:
|
||
vid_path = f"{vid_dir}/obj_{int(track_id):03d}.mp4"
|
||
vid_dict[track_id] = {"vid_path": vid_path, "vid_url": None}
|
||
except (ValueError, TypeError):
|
||
print(f"无效的track_id: {track_id}，跳过")
|
||
|
||
# 找到未上传的视频
|
||
unuploaded_vids = []
|
||
for track_id, vid_info in vid_dict.items():
|
||
if isinstance(vid_info, dict) and vid_info.get("vid_url") is None:
|
||
unuploaded_vids.append(vid_info["vid_path"])
|
||
|
||
# 上传未上传的视频并获取 URL
|
||
uploaded_urls = upload_files_and_get_urls_concurrently(
|
||
file_path_list=unuploaded_vids,
|
||
max_workers=8
|
||
)
|
||
|
||
# 更新视频URL
|
||
for track_id, vid_info in vid_dict.items():
|
||
if isinstance(vid_info, dict) and vid_info.get("vid_url") is None:
|
||
if vid_info["vid_path"] in unuploaded_vids:
|
||
idx = unuploaded_vids.index(vid_info["vid_path"])
|
||
if idx < len(uploaded_urls):
|
||
vid_info["vid_url"] = uploaded_urls[idx]
|
||
|
||
# 保存字典为 JSON 文件
|
||
with open(video_url_file, "w", encoding="utf-8") as f:
|
||
json.dump(vid_dict, f, ensure_ascii=False, indent=4)
|
||
|
||
return vid_dict
|
||
|
||
|
||
# 创建 Gradio 页面
|
||
with gr.Blocks() as demo:
|
||
gr.Markdown("# 📸 隐患排查系统 (sam3 + qwen3.5-27b)")
|
||
|
||
with gr.Row():
|
||
with gr.Column():
|
||
|
||
initial_files: list[str] = load_file_list()
|
||
default_video = initial_files[0] if initial_files else None
|
||
vid_file = gr.Dropdown(
|
||
label="视频",
|
||
choices=initial_files,
|
||
value=default_video, # 默认选中第一个
|
||
interactive=True,
|
||
)
|
||
reload_file_list_button = gr.Button("刷新视频列表")
|
||
reload_file_list_button.click(fn=reload_files, inputs=[], outputs=[vid_file])
|
||
|
||
run_sam3 = gr.Checkbox(label="1. 运行 SAM3 模型", value=True)
|
||
run_inspection = gr.Checkbox(label="2. 运行隐患排查", value=True)
|
||
gen_report = gr.Checkbox(label="3. 生成报告", value=True)
|
||
audio_recognition = gr.Checkbox(label="4. 运行音频识别", value=False)
|
||
|
||
run_button = gr.Button("运行", variant="primary")
|
||
|
||
get_vid_path_btn = gr.Button("获取视频路径")
|
||
full_path_text = gr.Textbox(visible=False)
|
||
|
||
with gr.Column():
|
||
preview = gr.Image(label="预览", scale=2)
|
||
img_slider = gr.Slider(label="帧索引", minimum=0, maximum=0, value=0, step=1)
|
||
textbox = gr.Textbox(label="隐患结果", lines=10)
|
||
jsonbox = gr.JSON(label="隐患结果json")
|
||
|
||
|
||
|
||
run_button.click(fn=run, inputs=[vid_file, run_sam3, run_inspection, gen_report], outputs=[img_slider, preview, textbox, jsonbox])
|
||
img_slider.change(fn=update_preview, inputs=[img_slider, vid_file], outputs=[preview, textbox], show_progress="hidden")
|
||
get_vid_path_btn.click(fn=get_full_vid_path, inputs=vid_file, outputs=full_path_text)
|
||
|
||
|
||
def get_class_tag_by_frame(data, idx, fps):
|
||
"""
|
||
根据给定的帧索引 (idx)，返回在该帧范围内的所有对象的 class:tag 信息。
|
||
|
||
参数:
|
||
data (dict): 包含 'tag', 'objects' 的字典数据（hazard_inspection.json格式）
|
||
idx (int): 需要查询的帧索引
|
||
|
||
返回:
|
||
str: 符合条件的 class:tag 列表，多个结果之间用换行符分隔。如果没有匹配项，返回空字符串。
|
||
"""
|
||
# 参数检查
|
||
if not isinstance(data, dict):
|
||
raise ValueError("数据必须是字典类型")
|
||
if 'tag' not in data or 'objects' not in data:
|
||
raise ValueError("数据必须包含 'tag' 和 'objects' 键")
|
||
|
||
tag_list = data['tag']
|
||
objects = data['objects']
|
||
|
||
interval = fps
|
||
idx = int(idx / interval) #转换
|
||
|
||
# 用于存储符合条件的 class:tag 字符串
|
||
result = []
|
||
all_class_tag = []
|
||
|
||
# 遍历每个物体
|
||
for obj in objects:
|
||
# 根据 tag_id 获取对应的隐患标签字符串
|
||
tag_id = obj.get('tag_id', 0)
|
||
class_id = obj.get('class_id', 0)
|
||
tag_str = tag_list[tag_id] if tag_id < len(tag_list) else f"未知标签({tag_id})"
|
||
location = obj.get('location', '')
|
||
start_frame = obj.get('start_frame', 0)
|
||
level = obj.get('level', '')
|
||
|
||
# 检查帧范围是否包含 idx
|
||
if start_frame == idx:
|
||
result.append(f"{tag_str} | 等级:{level} | 位置: {location}")
|
||
all_class_tag.append(f"{tag_str} | class_id:{class_id} | 等级:{level} | 开始帧:{start_frame} | 位置:{location}")
|
||
|
||
# 使用换行符连接所有结果
|
||
output = f"当前帧隐患：\n"+"\n".join(result)+"\n\n"+"所有隐患对象信息：\n"+"\n".join(all_class_tag)
|
||
return output
|
||
|
||
|
||
# 启动应用
|
||
if __name__ == "__main__":
|
||
demo.launch(
|
||
debug=True,
|
||
allowed_paths=[VIDEO_FOLDER]
|
||
) |