TensorFlow目标检测报错：无法从object_detection导入model_lib_v2

环境配置 #

• TensorFlow == 2.10.0
• Python 3.10.9
• 使用Jupyter Notebook，未启用TensorFlow GPU
• 模型：ssd_mobilenet_v2_fpnlite_320x320_coco17_tpu-8（TF2 20200711版本）

错误详情 #

执行以下训练命令时触发ImportError：

(odball) C:\Users\someone\Desktop\website\2023\TensorFlow Object Detection\TFODBall>python Tensorflow\models\research\object_detection\model_main_tf2.py --model_dir=Tensorflow\workspace\models\my_ssd_mobnet --pipeline_config_path=Tensorflow\workspace\models\my_ssd_mobnet\pipeline.config --num_train_steps=4000

错误日志：

Traceback (most recent call last):
  File "C:\Users\someone\Desktop\website\2023\TensorFlow Object Detection\TFODBall\Tensorflow\models\research\object_detection\model_main_tf2.py", line 31, in <module>
    from object_detection import model_lib_v2
ImportError: cannot import name 'model_lib_v2' from 'object_detection' (C:\Users\someone\Desktop\website\2023\TensorFlow Object Detection\TFODBall\odball\lib\site-packages\object_detection\__init__.py)

IDE提示无法解析object_detection导入。

已尝试的解决方法 #

• pip install tensorflow-object-detection-api
• 卸载并重装TensorFlow
• 重启Jupyter内核

当前环境pip包列表 #

解决方案 #

1. 修正Python路径优先级 #

错误显示Python优先使用site-packages中的object_detection包，而非本地克隆的TensorFlow Models库。需将本地库路径加入环境变量：

• 命令行训练前执行：

  set PYTHONPATH=C:\Users\someone\Desktop\website\2023\TensorFlow Object Detection\TFODBall\Tensorflow\models\research;C:\Users\someone\Desktop\website\2023\TensorFlow Object Detection\TFODBall\Tensorflow\models\research\slim
  

• Jupyter Notebook中添加：

  import sys
  sys.path.insert(0, 'C:/Users/someone/Desktop/website/2023/TensorFlow Object Detection/TFODBall/Tensorflow/models/research')
  sys.path.insert(0, 'C:/Users/someone/Desktop/website/2023/TensorFlow Object Detection/TFODBall/Tensorflow/models/research/slim')
  

2. 编译Protobuf文件 #