本文介绍了两套基于YOLOv8模型的活体检测服务系统。
1️⃣：一种针对图片YOLO活体检测服务【端口：5000】
- ① 脚本
import os import json import threading from flask import Flask, request, jsonify from ultralytics import YOLO app = Flask(__name__) class YOLOService: def __init__(self): self.model = None self.lock = threading.Lock() self.load_model() def load_model(self): """加载能检测人和车的模型""" model_path = "/opt/yolov11/models/yolov8n.pt" print(f"【重要】加载新模型: {model_path}") print(f"此模型可检测80类目标，包括人和车") try: if os.path.exists(model_path): self.model = YOLO(model_path) # 打印模型能识别的类别 if hasattr(self.model, 'names'): print(f"模型可识别类别: {self.model.names}") # 找到人和车的ID (COCO数据集中: 0=person, 2=car) print(f"-> 人的类别ID: 0 (person)") print(f"-> 车的类别ID: 2 (car)") print("✅ 模型加载成功!") else: print(f"❌ 错误: 模型文件不存在 {model_path}") print("请运行: cd /opt/yolov11/models && wget -O yolov8n.pt https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n.pt") self.model = None except Exception as e: print(f"❌ 模型加载失败: {e}") self.model = None def predict_image(self, image_path): """执行推理，只返回人和车""" if self.model is None: return {"error": "Model not loaded", "success": False} try: with self.lock: # 执行推理 results = self.model(image_path, verbose=False) # 提取检测结果 all_detections = [] person_count = 0 car_count = 0 for result in results: for box in result.boxes: cls_id = int(box.cls) conf = float(box.conf) # 只保留人和车 (COCO数据集中: 0=person, 2=car) if cls_id in [0, 2]: class_name = "person" if cls_id == 0 else "car" if cls_id == 0: person_count += 1 else: car_count += 1 all_detections.append({ "class": cls_id, "class_name": class_name, "confidence": conf, "bbox": box.xyxy[0].tolist() # [x1, y1, x2, y2] }) return { "success": True, "detections": all_detections, "statistics": { "total_objects": len(all_detections), "person_count": person_count, "car_count": car_count }, "image_size": result.orig_shape } except Exception as e: return {"error": str(e), "success": False} service = YOLOService() @app.route('/health', methods=['GET']) def health_check(): model_status = "loaded" if service.model is not None else "not loaded" return jsonify({ "status": "healthy", "model_loaded": service.model is not None, "model_name": "yolov8n.pt", "model_path": "/opt/yolov11/models/yolov8n.pt", "detection_classes": "person (人), car (车)" }) @app.route('/predict', methods=['POST']) def predict(): """接收图片路径进行预测""" data = request.json if not data or 'image_path' not in data: return jsonify({"error": "Missing image_path parameter", "success": False}), 400 image_path = data['image_path'] # 验证文件是否存在 if not os.path.exists(image_path): return jsonify({"error": f"File not found: {image_path}", "success": False}), 404 # 执行预测 result = service.predict_image(image_path) return jsonify(result) @app.route('/predict/url', methods=['POST']) def predict_url(): """通过URL推理""" data = request.json if not data or 'image_url' not in data: return jsonify({"error": "Missing image_url parameter", "success": False}), 400 image_url = data['image_url'] result = service.predict_image(image_url) return jsonify(result) if __name__ == '__main__': port = 5000 print(f"✅ YOLO活体检测服务启动!") print(f" 端口: {port}") print(f" 模型: yolov8n.pt (专检测人和车)") print(f" 访问: http://localhost:{port}/health") app.run(host='0.0.0.0', port=port, threaded=True, debug=False)- ② 服务
[Unit] Description=YOLOv11 CPU Inference Service After=network.target [Service] Type=simple User=zst Group=zst WorkingDirectory=/opt/yolov11 Environment="PATH=/home/zst/yolov11_venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin" ExecStart=/home/zst/yolov11_venv/bin/python /opt/yolov11/api/inference_service.py Restart=on-failure RestartSec=5 StandardOutput=journal StandardError=journal SyslogIdentifier=yolov11-service [Install] WantedBy=multi-user.target2️⃣：一种针对视频YOLO活体检测服务【端口：5001】【以下任选一种脚本】
- ① 脚本
import os import cv2 import threading import time import json import traceback from flask import Flask, request, jsonify from ultralytics import YOLO app = Flask(__name__) class VideoDetectionService: def __init__(self): self.model = None self.lock = threading.Lock() self.load_model() def load_model(self): """加载正确的模型 - yolov8n.pt""" # 【重要】使用正确的模型路径 model_path = "/opt/yolov11/models/yolov8n.pt" print(f"【视频服务】加载模型: {model_path}") # 检查文件是否存在 if not os.path.exists(model_path): print(f"【视频服务】❌ 错误: 模型文件不存在!") print(f"请运行: wget -O {model_path} https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n.pt") self.model = None return try: self.model = YOLO(model_path) print(f"【视频服务】✅ 模型加载成功!") print(f"【视频服务】模型信息: {len(self.model.names)}个类别") # 验证模型能检测人和车 if hasattr(self.model, 'names'): print(f"【视频服务】人(person): ID 0") print(f"【视频服务】车(car): ID 2")