Windows下pip安装tf-object-detection遇参数错误与依赖冲突求助

我在Windows上装TensorFlow Object Detection时碰到俩问题，跟论坛里同标题的错误都不一样，下面说下问题和解决步骤：

第一个坑：--use-feature=2020-resolver 参数无效 #

我执行了这条命令：

C:\Users\NAME\models\research>python -m pip install --use-feature=2020-resolver .

结果直接报错：

C:\Users\NAME\models\research>python -m pip install --use-feature=2020-resolver .

Usage:
  C:\Users\NAME\AppData\Local\Programs\Python\Python311\python.exe -m pip install [options] <requirement specifier> [package-index-options] ...
  C:\Users\NAME\AppData\Local\Programs\Python\Python311\python.exe -m pip install [options] -r <requirements file> [package-index-options] ...
  C:\Users\NAME\AppData\Local\Programs\Python\Python311\python.exe -m pip install [options] [-e] <vcs project url> ...
  C:\Users\NAME\AppData\Local\Programs\Python\Python311\python.exe -m pip install [options] [-e] <local project path> ...
  C:\Users\NAME\AppData\Local\Programs\Python\Python311\python.exe -m pip install [options] <archive url/path> ...

option --use-feature: invalid choice: '2020-resolver' (choose from 'fast-deps', 'truststore', 'no-binary-enable-wheel-cache')

原因很简单：新版pip（21.0及以上）默认就用2020版依赖解析器了，这个--use-feature=2020-resolver参数已经被删掉了，所以用了就会报错。

第二个坑：去掉参数后出现依赖冲突 #

我把参数删掉，执行python -m pip install .，结果又炸了，报依赖冲突：

ERROR: Cannot install object-detection and object-detection==0.1 because these package versions have conflicting dependencies.

The conflict is caused by:
    tf-models-official 2.11.3 depends on tensorflow-addons
    tf-models-official 2.11.2 depends on tensorflow-addons
    tf-models-official 2.11.0 depends on opencv-python-headless==4.5.2.52
    object-detection 0.1 depends on sacrebleu<=2.2.0
    tf-models-official 2.10.1 depends on sacrebleu==2.2.0
    tf-models-official 2.10.0 depends on opencv-python-headless==4.5.2.52
    tf-models-official 2.9.2 depends on tensorflow-addons
    tf-models-official 2.9.1 depends on tensorflow-addons
    tf-models-official 2.9.0 depends on tensorflow-addons
    tf-models-official 2.8.0 depends on tensorflow-addons
    tf-models-official 2.7.2 depends on tensorflow-addons
    tf-models-official 2.7.1 depends on tensorflow-addons
    tf-models-official 2.7.0 depends on tensorflow-addons
    tf-models-official 2.6.1 depends on tensorflow-addons
    tf-models-official 2.6.0 depends on tensorflow-addons
    tf-models-official 2.5.1 depends on tensorflow-addons

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict

ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts

我当前环境的pip freeze输出是：

absl-py==1.4.0
astunparse==1.6.3
cachetools==5.3.0
certifi==2022.12.7
charset-normalizer==3.1.0
contextlib2==21.6.0
contourpy==1.0.7
cycler==0.11.0
Cython==0.29.33
flatbuffers==23.3.3
fonttools==4.39.2
gast==0.4.0
google-auth==2.16.2
google-auth-oauthlib==0.4.6
google-pasta==0.2.0
grpcio==1.51.3
h5py==3.8.0
idna==3.4
jax==0.4.6
keras==2.12.0rc1
kiwisolver==1.4.4
libclang==15.0.6.1
lvis==0.5.3
lxml==4.9.2
Markdown==3.4.1
MarkupSafe==2.1.2
matplotlib==3.7.1
numpy==1.23.5
oauthlib==3.2.2
object-detection @ file:///C:/Users/NAME/models/research
opencv-python==4.7.0.72
opt-einsum==3.3.0
packaging==23.0
pandas==1.5.3
Pillow==9.4.0
protobuf==4.22.1
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycocotools==2.0.6
pygame==2.1.3.dev8
pyparsing==3.0.9
python-dateutil==2.8.2
pytz==2022.7.1
requests==2.28.2
requests-oauthlib==1.3.1
rsa==4.9
scipy==1.10.1
six==1.16.0
tensorboard==2.12.0
tensorboard-data-server==0.7.0
tensorboard-plugin-wit==1.8.1
tensorflow==2.12.0rc1
tensorflow-estimator==2.12.0rc0
tensorflow-intel==2.12.0rc1
tensorflow-io-gcs-filesystem==0.31.0
termcolor==2.2.0
tf-slim==1.1.0
typing_extensions==4.5.0
urllib3==1.26.15
Werkzeug==2.2.3
wrapt==1.14.1

一步步解决的方法 #

1. 先清掉冲突的包 #

先把已经装的object-detection和可能搞事情的tf-models-official卸载：

pip uninstall -y object-detection tf-models-official

2. 降级TensorFlow到稳定版 #

你现在用的是TensorFlow 2.12.0rc1（预发布版），这种版本最容易出依赖问题，换成稳定的2.11.0：

pip install tensorflow==2.11.0 tensorflow-estimator==2.11.0

3. 解决OpenCV版本冲突 #

你当前装的是opencv-python==4.7.0.72，但tf-models-official要求的是opencv-python-headless==4.5.2.52，先卸载旧的，装指定版本：

pip uninstall -y opencv-python
pip install opencv-python-headless==4.5.2.52

4. 装tensorflow-addons #

错误提示里说tf-models-official依赖这个包，先装上：

pip install tensorflow-addons

5. 安装本地的object-detection包 #

回到models/research目录，用--no-deps参数跳过自动依赖解析，避免再次触发冲突：

cd C:\Users\NAME\models\research
python -m pip install --no-deps .

6. 验证是否装好了 #

跑个简单的命令测试下：

python -c "import object_detection; print('安装成功')"

怎么看本地包的依赖列表？ #

要看models/research里的object-detection依赖啥，直接看该目录下的setup.py文件，找install_requires字段，里面列的就是所有依赖；或者在该目录下执行：

pip show -f object-detection

如果还没安装，也可以用pip install --dry-run .来预览它要装的依赖。

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