Colab运行TF2目标检测model_main_tf2.py无法导入model_lib_v2如何解决

• 第一步：卸载全局安装的旧版object_detection包
  报错显示Python优先读取了全局dist-packages目录下的object_detection，该版本无model_lib_v2模块，先执行命令卸载：
  pip uninstall -y object-detection
• 第二步：调整Python模块查找优先级
  需将本地存放TF Models的research、research/slim目录加到系统路径最前，避免优先读取全局包：
  命令行运行脚本前执行：

  export PYTHONPATH="/content/drive/MyDrive/Colab_Notebooks/RealTimeObjectDetection/Tensorflow/models/research:/content/drive/MyDrive/Colab_Notebooks/RealTimeObjectDetection/Tensorflow/models/research/slim:$PYTHONPATH"
  

  Colab环境下可在代码块开头添加：

  import sys
  sys.path.insert(0, "/content/drive/MyDrive/Colab_Notebooks/RealTimeObjectDetection/Tensorflow/models/research")
  sys.path.insert(0, "/content/drive/MyDrive/Colab_Notebooks/RealTimeObjectDetection/Tensorflow/models/research/slim")
  

• 第三步：验证本地TF Models文件完整性
  进入/content/drive/MyDrive/Colab_Notebooks/RealTimeObjectDetection/Tensorflow/models/research/object_detection目录，确认存在model_lib_v2.py文件，若不存在说明拉取的Models版本过旧，拉取master分支的最新代码即可。
• 第四步：重新本地安装object_detection包
  在research目录下执行以下命令，用本地TF2适配的配置安装：

  cp object_detection/packages/tf2/setup.py .
  python -m pip install .
  

内容的提问来源于stack exchange，提问作者Ninomiya Hatsu