【笔记】pip和cmake编译安装opencv的文件对比-Windows和Linux

谁践踏了优雅 2022-12-01 03:47 344阅读 0赞

# cmake编译时的配置： #

注意Python3 install path：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 1][]

注意把TESTS去掉，大概有37个工程![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 2][]

# Linux下cmake编译安装和pip安装opencv-python的区别 #

## cmake编译安装opencv-python ##

编译安装后，出现6个文件，

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 3][]

在python-3.7文件下，出现.so文件：

![20200825174554296.PNG][]

config-3.7.py文件下，配置的是.so文件的配置路径：

PYTHON_EXTENSIONS_PATHS = [
        os.path.join('/home/pytorch/anaconda3/lib/python3.7/site-packages/cv2', 'python-3.7')
    ] + PYTHON_EXTENSIONS_PATHS

config.py文件下，配置的是.so文件引用的二进制文件的路径：

import os
    
    BINARIES_PATHS = [
        os.path.join('/usr/local/opencv440', 'lib')
    ] + BINARIES_PATHS

如果二进制文件配置在系统环境变量里，该文件应该就没作用了。

如果移植到别的电脑，只要配置这些文件即可。

## pip安装OpenCV-python ##

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 4][]

data文件夹下的内容：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 5][]

opencv3.4和opencv4.x版本的contrib是不一样的，编译安装时，注意选择对应版本

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 6][]

# Windows下VS 编译安装支持CUDA的opencv-python #

## 手动编译安装opencv-python ##

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 7][]

config.py文件，配置CUDA环境和二进制文件的编译

import os
    
    BINARIES_PATHS = [
        os.path.join('C:/software/opencv440/VS2019-CUDA-Python/install', 'x64/vc16/bin'),
        os.path.join(os.getenv('CUDA_PATH', 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.2'), 'bin')
    ] + BINARIES_PATHS

config-3.8.py文件，配置cv2的路径：

PYTHON_EXTENSIONS_PATHS = [
        os.path.join('C:/ProgramData/Anaconda3/Lib/site-packages/cv2', 'python-3.8')
    ] + PYTHON_EXTENSIONS_PATHS

python-3.8文件夹中的文件：

![20200826090650648.PNG][]

## pip 安装opencv-python ##

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 8][]

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70]: /images/20221123/abcc84eca31f463d9836f47b2c94b2f5.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 1]: /images/20221123/0f7b981db76e4624b0a1687b1b9c4f28.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 2]: /images/20221123/3e15960b26ae4cba8f3d1b2f3fd9a404.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 3]: /images/20221123/357af94d218c438ea57449f1bdc985b0.png
[20200825174554296.PNG]: /images/20221123/6ba97881405b49b091c8f28c188d8879.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 4]: /images/20221123/8b65a3f2d56045978cf109d66cb0d9ea.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 5]: /images/20221123/d42d62791db14d8c81973c7d184f8078.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 6]: /images/20221123/883988bb5c3f4294a7bdf2cc8cc40d34.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 7]: /images/20221123/dd65550e58e649ff80b71307f4ca4314.png
[20200826090650648.PNG]: /images/20221123/dbc52efb115242f5ade5f6d3e41b78b8.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 8]: /images/20221123/ec086eaf94054d7bb30784762d94e0ff.png