1.初始化仓库\n2.添加同时推流mp4文件脚本\n3.添加批量添加摄像头

001测试模型.py

@@ -0,0 +1,51 @@
+from ultralytics import YOLO
+import cv2
+
+def predict_and_visualize(model_path, image_path, output_path):
+    # 加载训练好的模型
+    model = YOLO(model_path)
+    
+    # 进行预测
+    results = model.predict(source=image_path, conf=0.25)  # conf设置置信度阈值
+    
+    # 读取原始图片
+    img = cv2.imread(image_path)
+    
+    # 获取预测结果
+    boxes = results[0].boxes
+    class_names = model.names  # 获取类别名称字典
+    
+    # 遍历每个检测结果
+    for box in boxes:
+        # 获取坐标和类别信息
+        x1, y1, x2, y2 = map(int, box.xyxy[0].tolist())
+        cls_id = int(box.cls[0].item())
+        conf = box.conf[0].item()
+        
+        # 绘制边界框
+        color = (0, 255, 0)  # 绿色边框
+        cv2.rectangle(img, (x1, y1), (x2, y2), color, 2)
+        
+        # 准备显示文本
+        label = f"{class_names[cls_id]}: {conf:.2f}"
+        
+        # 计算文本位置
+        (w, h), _ = cv2.getTextSize(label, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 1)
+        
+        # 绘制文本背景
+        cv2.rectangle(img, (x1, y1 - h - 5), (x1 + w, y1), color, -1)
+        # 绘制文本
+        cv2.putText(img, label, (x1, y1 - 5), 
+                   cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 0), 1)
+    
+    # 保存结果
+    cv2.imwrite(output_path, img)
+    print(f"结果已保存至: {output_path}")
+
+if __name__ == "__main__":
+    # 使用示例
+    model_path = "models/安全帽检测模型/yolo11n_safehat.pt"  # 替换为你的模型路径
+    image_path = "images/mp4_509.jpg"  # 替换为你的图片路径
+    output_path = "output/mp4_509.jpg"    # 输出文件名
+    
+    predict_and_visualize(model_path, image_path, output_path)

002批量测试模型图片.py

@@ -0,0 +1,63 @@
+from ultralytics import YOLO
+import cv2
+import os 
+
+def predict_and_visualize(model_path, image_path, output_path):
+    # 加载训练好的模型
+    model = YOLO(model_path)
+    
+    all_image_file = os.listdir(image_path)
+    all_image_path = [os.path.join(image_path, t) for t in all_image_file]
+    
+    for i in range(len(all_image_path)):
+    
+        # 进行预测
+        results = model.predict(source=all_image_path[i], conf=0.25)  # conf设置置信度阈值
+        
+        # 读取原始图片
+        img = cv2.imread(all_image_path[i])
+        
+        # 获取预测结果
+        boxes = results[0].boxes
+        class_names = model.names  # 获取类别名称字典
+        
+        
+        # {0: 'head', 1: 'safehat'}
+        print(class_names)
+        
+        # break
+        
+        # 遍历每个检测结果
+        for box in boxes:
+            # 获取坐标和类别信息
+            x1, y1, x2, y2 = map(int, box.xyxy[0].tolist())
+            cls_id = int(box.cls[0].item())
+            conf = box.conf[0].item()
+            
+            # 绘制边界框
+            color = (0, 255, 0)  # 绿色边框
+            cv2.rectangle(img, (x1, y1), (x2, y2), color, 2)
+            
+            # 准备显示文本
+            label = f"{class_names[cls_id]}: {conf:.2f}"
+            
+            # 计算文本位置
+            (w, h), _ = cv2.getTextSize(label, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 1)
+            
+            # 绘制文本背景
+            cv2.rectangle(img, (x1, y1 - h - 5), (x1 + w, y1), color, -1)
+            # 绘制文本
+            cv2.putText(img, label, (x1, y1 - 5), 
+                    cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 0), 1)
+        
+        # 保存结果
+        cv2.imwrite(output_path+f"{i}.jpg", img)
+        print(f"结果已保存至: {output_path}")
+
+if __name__ == "__main__":
+    # 使用示例
+    model_path = "/home/admin-root/haotian/xcms/models/安全帽检测模型OpenVINO/best_s.xml"  # 替换为你的模型路径
+    image_path = "/home/admin-root/haotian/锻8/tensorrtx/yolov8/images"  # 替换为你的图片路径
+    output_path = "/home/admin-root/haotian/xcms/output/"    # 输出文件名
+    
+    predict_and_visualize(model_path, image_path, output_path)

003推流指定文件夹下所有mp4文件.sh

@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+
+VIDEO_DIR="/home/admin-root/haotian/康达瑞贝斯机器狗/data_video"  # 你的 .mp4 文件目录
+RTSP_SERVER="rtsp://10.0.0.17:8554/camera_test"
+
+
+# 捕获 SIGINT（Ctrl+C）和 SIGTERM，触发 exit（进而触发 EXIT trap）
+trap "exit" INT TERM
+
+# EXIT 触发时，终止当前进程组（包括所有子进程）
+trap "kill 0" EXIT
+
+for filepath in "$VIDEO_DIR"/*.mp4; do
+  [ -e "$filepath" ] || continue
+  filename=$(basename "$filepath" .mp4)
+  RTSP_URL="${RTSP_SERVER}/${filename}"
+
+  (
+    while true; do
+      echo "$(date): 推流 -> $RTSP_URL"
+      ffmpeg -re -stream_loop -1 -i "$filepath" \
+        -vf "scale=1920:1080" \
+        -c:v libx264 -preset veryfast -tune zerolatency \
+        -c:a aac -b:a 128k \
+        -muxdelay 0 -muxpreload 0 \
+        -f rtsp -rtsp_transport tcp \
+        "$RTSP_URL"
+      echo "$(date): 推流中断，5 秒后重启 -> $RTSP_URL"
+      sleep 5
+    done
+  ) &
+done
+
+# 等待所有后台任务
+wait

004批量添加摄像头.py

@@ -0,0 +1,61 @@
+import requests 
+import uuid
+import os
+import time
+
+"""
+    批量向xcms中添加添加摄像头
+"""
+
+RSTP_URL = "rtsp://10.0.0.17:8554/camera_test/"
+ADD_CAMERA_URL = "http://10.0.0.81:9001/open/addStream"
+def add_camera(rtsp_path, nick_name, code ,method="POST"):
+    
+    # code = uuid.uuid4().hex
+    
+    # print(code)
+    
+    data = {
+        "code": f"{code}",# 必填，摄像头编号（不能与已有摄像头编号重复）
+        "nickname": f"{nick_name}",#必填，名称
+        "pull_stream_type": 1, # 必填，数值类型，1:RTSP,2:RTMP,3:FVL,4:HLS,21:GB28181
+        "pull_stream_url": f"{rtsp_path}", # 必填，直播流地址
+        "pull_stream_ip":"10.0.0.17", # 必填，摄像头IP
+        "pull_stream_port":8554, # 必填，数值类型，摄像头拉流服务对用的端口
+        # "camera_name":"", # 非必填，摄像头名称
+        # "camera_manufacturer":"", # 非必填，摄像头厂商
+        # "camera_device_id":"group1",# （v4.638新增）非必填，摄像头所属分组编号，默认不填写时等于code
+        # "remark":"", # 非必填，备注
+        # "onvif_username":"", # 非必填，探测ONVIF username
+        # "onvif_password":"", # 非必填，探测ONVIF password
+        "is_audio":0 # 必填，数值类型，0:静音 1:原始音频
+    }
+    
+    headers = {
+        "Content-Type": "application/json",
+        "Safe":"aqxY9ps21fyhyKNRyYpGvJCTp1JBeGOM"
+    }
+    
+    response = requests.post(ADD_CAMERA_URL, json=data, headers=headers)
+    
+    response.raise_for_status()
+    
+    response_data = response.json()
+    
+    print(response_data["code"])
+    
+
+def main():
+    
+    video_path = "/home/admin-root/haotian/康达瑞贝斯机器狗/data_video"
+    video_name_list = os.listdir(video_path)
+    
+    for i, video_name in enumerate(video_name_list):
+        video_name = video_name.split(".")[0]
+        add_camera(RSTP_URL+video_name, video_name, f"t{i}")
+        time.sleep(0.2)
+        
+        
+        
+if __name__ == "__main__":
+    main()

docs/模型测试结果.md

@@ -0,0 +1,15 @@
+## 模型可用性
+
+- [X]  人脸检测模型
+- [X]  厨师帽检测模型
+- [ ]  反光衣检测模型
+  AttributeError: Can't get attribute 'C3k2' on <module 'ultralytics.nn.modules.block' from '/home/admin-root/miniconda3/envs/trt/lib/python3.9/site-packages/ultralytics/nn/modules/block.py'>
+- [X]  安全帽检测模型
+  报错同上.
+- [X]  打架检测模型
+- [X]  抽烟检测模型
+- [X]  持械检测模型
+- [X]  火焰烟火检测模型
+- [X]  烟尘检测模型
+- [X]  睡岗检测模型
+

models/安全帽检测模型OpenVINO/test.py

@@ -0,0 +1,16 @@
+from ultralytics import YOLO
+import openvino as ov
+'''
+    加载文件失败
+'''
+# 加载 OpenVINO 模型
+core = ov.Core()
+compiled_model = core.compile_model("best_s.xml", "CPU")
+
+# 通过 YOLO 包装推理
+model = YOLO("best_s.xml")  # 使用 YOLO 接口
+model.model = compiled_model  # 替换为 OpenVINO 模型
+
+
+# 执行推理
+results = model.predict("image.jpg")