下载预训练模型的时间取决于**网络速度**和**模型文件大小**，以下是详细分析和优化建议： --- ### **1. 模型下载时间估算** - **YOLOv5s.onnx**（常用车辆检测模型）： - 文件大小：约 **28MB**（压缩包解压后可能更大） - 下载时间参考： | 网络带宽 | 下载时间 | |----------|----------| | 10Mbps | 3-5分钟 | | 50Mbps | 30-60秒 | | 100Mbps+ | <30秒 | - **其他常见模型**： - MobileNet-SSD（Caffe模型）：约 **20MB** - Faster R-CNN（PyTorch转ONNX）：约 **100-200MB** --- ### **2. 加速下载的方法** #### **方法1：使用国内镜像源** ```python # 修改下载代码，使用清华/中科大镜像 def _download_default_model(self, save_path): mirrors = [ "https://mirrors.tuna.tsinghua.edu.cn/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx", "https://mirrors.ustc.edu.cn/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx" ] for url in mirrors: try: print(f"尝试从 {url} 下载...") urllib.request.urlretrieve(url, save_path) return except Exception as e: print(f"镜像 {url} 下载失败: {str(e)}") raise RuntimeError("所有镜像下载失败，请检查网络") ``` #### **方法2：手动下载后放置到指定路径** 1. 直接访问模型下载地址： - 官方源：`https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx` - 国内镜像：`https://mirror.ghproxy.com/https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx` 2. 下载后保存到脚本同目录，命名为 `yolov5s.onnx`。 #### **方法3：使用多线程下载工具** ```python # 使用requests库实现断点续传（需安装：pip install requests） def _download_with_progress(url, save_path): import requests from tqdm import tqdm # 进度条库 response = requests.get(url, stream=True) total_size = int(response.headers.get('content-length', 0)) block_size = 1024 # 1KB with open(save_path, 'wb') as f, tqdm( desc=f"下载 {os.path.basename(save_path)}", total=total_size, unit='iB', unit_scale=True ) as bar: for data in response.iter_content(block_size): f.write(data) bar.update(len(data)) ``` --- ### **3. 下载状态检查** #### **实时进度显示** ```python def _download_default_model(self, save_path): url = "https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx" try: print("开始下载模型...") # 使用urllib的进度回调（Python 3.8+） def _progress_hook(count, block_size, total_size): percent = count * block_size / total_size * 100 print(f"\r下载进度: {percent:.1f}%", end="") urllib.request.urlretrieve(url, save_path, reporthook=_progress_hook) print("\n下载完成！") except Exception as e: raise RuntimeError(f"下载失败: {str(e)}") ``` #### **断点续传实现** ```python def _download_with_resume(url, save_path): import os if os.path.exists(save_path): mode = 'ab' # 追加模式 offset = os.path.getsize(save_path) else: mode = 'wb' offset = 0 req = urllib.request.Request(url, headers={'Range': f'bytes={offset}-'}) with open(save_path, mode) as f: with urllib.request.urlopen(req) as response: while True: chunk = response.read(8192) if not chunk: break f.write(chunk) ``` --- ### **4. 完整修复后的代码** ```python import os import urllib.request import sys class VehicleDetector: def __init__(self, model_path="yolov5s.onnx"): if not os.path.exists(model_path): self._download_default_model(model_path) try: import cv2 self.net = cv2.dnn.readNetFromONNX(model_path) print("模型加载成功！") except Exception as e: raise RuntimeError(f"模型加载失败: {str(e)}") def _download_default_model(self, save_path): mirrors = [ "https://mirror.ghproxy.com/https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx", "https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5s.onnx" ] for url in mirrors: try: print(f"从 {url} 下载模型...") self._download_with_progress(url, save_path) return except Exception as e: print(f"下载失败: {str(e)}，尝试下一个镜像...") raise RuntimeError("所有镜像下载失败，请检查网络或手动下载") def _download_with_progress(self, url, save_path): def _progress_hook(count, block_size, total_size): percent = count * block_size / total_size * 100 print(f"\r下载进度: {percent:.1f}%", end="") urllib.request.urlretrieve(url, save_path, reporthook=_progress_hook) print("\n下载完成！") # 测试代码 if __name__ == "__main__": try: detector = VehicleDetector() except Exception as e: print(f"初始化失败: {e}") sys.exit(1) ``` --- ### **5. 常见问题解决** #### **Q1: 下载速度慢怎么办？** - 使用国内镜像（如清华源、中科大源） - 尝试在非高峰时段下载 - 使用IDM等多线程下载工具 #### **Q2: 下载中断后如何继续？** - 代码中已实现断点续传逻辑 - 或手动删除不完整的文件后重新运行 #### **Q3: 下载后模型仍无法加载？** - 检查文件大小是否完整（yolov5s.onnx应约28MB） - 验证OpenCV版本： ```python import cv2 print(cv2.__version__) # 需要≥4.5.1 ``` ---