Windows11下detectron2与LayoutParser安装报错的解决求助

问题描述 #

1. 初始安装尝试与Detectron2未找到错误 #

执行以下命令安装相关工具：

pip install -U layoutparser
pip install 'git+https://github.com/facebookresearch/detectron2.git@v0.4#egg=detectron2'
pip install layoutparser[ocr]
git clone https://github.com/Layout-Parser/layout-parser.git

运行LayoutParser模型初始化代码：

model = lp.models.Detectron2LayoutModel('lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config', 
                                 extra_config=["MODEL.ROI_HEADS.SCORE_THRESH_TEST", 0.8],
                                 label_map={0: "Text", 1: "Title", 2: "List", 3:"Table", 4:"Figure"})

返回错误：

ImportError:
Detectron2LayoutModel requires the detectron2 library but it was not found in your environment. Checkout the instructions on the
installation page: https://github.com/facebookresearch/detectron2/blob/master/INSTALL.md and follow the ones
that match your environment. Typically the following would work for MacOS or Linux CPU machines:
pip install 'git+https://github.com/facebookresearch/detectron2.git@v0.4#egg=detectron2'

2. Windows版Detectron2安装错误 #

尝试安装Windows适配版Detectron2时执行命令：

git clone https://github.com/DGMaxime/detectron2-windows.git
cd detectron2-windows
pip install -e 

收到错误：

Note: you may need to restart the kernel to use updated packages.
Usage:
C:\Users\msubh\anaconda3\envs\expOPlay\python.exe -m pip install [options] [package-index-options] ...
C:\Users\msubh\anaconda3\envs\expOPlay\python.exe -m pip install [options] -r [package-index-options] ...
C:\Users\msubh\anaconda3\envs\expOPlay\python.exe -m pip install [options] [-e] ...
C:\Users\msubh\anaconda3\envs\expOPlay\python.exe -m pip install [options] [-e] ...
C:\Users\msubh\anaconda3\envs\expOPlay\python.exe -m pip install [options] <archive url/path> ...

-e option requires 1 argument

3. 新环境下的依赖冲突问题 #

创建新环境执行pip install torch后，出现大量依赖冲突报错，例如：

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
gdown 4.7.1 requires requests[socks], which is not installed.
...（省略后续冲突信息）

安装requests后仍存在大量未解决的依赖冲突。

解决方案 #

1. 创建纯净conda环境 #

避免现有包干扰，创建并激活指定Python版本的环境：

conda create -n layoutparser_env python=3.8 -y
conda activate layoutparser_env

选择Python3.8是因为该版本与Detectron2、LayoutParser兼容性较好。

2. 安装适配Windows的PyTorch #

使用conda安装PyTorch，自动处理依赖问题：

# CPU版本
conda install pytorch torchvision torchaudio cpuonly -c pytorch -y

# 若有NVIDIA GPU，安装对应CUDA版本（示例为CUDA11.8）
# conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia -y

3. 正确安装Windows版Detectron2 #

修复pip install -e的参数缺失问题，完整命令如下：

git clone https://github.com/DGMaxime/detectron2-windows.git
cd detectron2-windows
pip install -e .

注意-e后必须跟.，代表安装当前目录下的项目。

4. 安装LayoutParser及OCR依赖 #

pip install -U layoutparser
pip install layoutparser[ocr]

5. 验证安装 #

运行以下代码测试模型初始化：

import layoutparser as lp
model = lp.models.Detectron2LayoutModel(
    'lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config', 
    extra_config=["MODEL.ROI_HEADS.SCORE_THRESH_TEST", 0.8],
    label_map={0: "Text", 1: "Title", 2: "List", 3:"Table", 4:"Figure"}
)

问题根源说明 #

• 官方Detectron2的pip安装命令仅适配Linux/Mac，Windows系统需使用专门适配的分支版本
• pip install -e命令需要指定安装路径，之前缺失.导致参数错误
• 直接用pip安装PyTorch易引发依赖冲突，conda能更好地管理跨平台依赖

内容的提问来源于stack exchange，提问作者DL_DS_2