在 msys2@mingw 下编译 BVLC/Caffe

原创

已于 2023-10-19 18:23:25 修改 · 786 阅读

1 ·

CC 4.0 BY-SA版权

文章标签：

#caffe #人工智能 #深度学习

于 2023-09-28 09:49:47 首次发布

本文介绍了在Windows上构建Caffe开发环境的步骤，包括所需工具安装、CMake配置以及遇到的常见错误如OpenCV版本兼容性、链接错误等的解决方法。

(参考 https://github.com/lemonsqueeze/mingw-caffe ，有关这个项目的尝试写在文章后面)

1. 准备开发环境

有关需要安装的工具，上面这篇已经说的很明白了，照搬于下。

pacman -S --needed git make patch diffutils
pacman -S --needed \
		mingw-w64-${MSYSTEM_CARCH}-cmake \
		mingw-w64-${MSYSTEM_CARCH}-python \
		mingw-w64-${MSYSTEM_CARCH}-tools-git \
		mingw-w64-${MSYSTEM_CARCH}-gcc \
		mingw-w64-${MSYSTEM_CARCH}-boost \
		mingw-w64-${MSYSTEM_CARCH}-protobuf-c \
		mingw-w64-${MSYSTEM_CARCH}-gflags \
		mingw-w64-${MSYSTEM_CARCH}-glog \
		mingw-w64-${MSYSTEM_CARCH}-hdf5 \
		mingw-w64-${MSYSTEM_CARCH}-openblas \
		mingw-w64-${MSYSTEM_CARCH}-leveldb \
		mingw-w64-${MSYSTEM_CARCH}-lmdb \
		mingw-w64-${MSYSTEM_CARCH}-snappy \
		mingw-w64-${MSYSTEM_CARCH}-python-matplotlib \
		mingw-w64-${MSYSTEM_CARCH}-python-pytest \
		mingw-w64-${MSYSTEM_CARCH}-python-scipy

还要下载 caffe 源代码

wget https://github.com/BVLC/caffe/archive/refs/tags/1.0.tar.gz
tar xf 1.0.tar.gz
cd caffe

2. 编译脚本

#!/bin/bash
 
BUILD_DIR=${1:-build}
export  PATH=/mingw64/bin:/mingw64/include:$PATH
 
[ -e ${BUILD_DIR} ] && rm -r ${BUILD_DIR}/* || mkdir ${BUILD_DIR} 
 
cmake \
	-B${BUILD_DIR} \
	-DBLAS=open \
	caffe-1.0
 
# Build
[ $? -eq 0 ] &&  cmake --build ${BUILD_DIR} -j 8

3. 错误修复

Error 1:

/home/rd/NN/caffe/src/caffe/layers/window_data_layer.cpp: In member function ‘virtual void caffe::WindowDataLayer<Dtype>::load_batch(caffe::Batch<Dtype>*)’:
/home/rd/NN/caffe/src/caffe/layers/window_data_layer.cpp:293:42: error: ‘CV_LOAD_IMAGE_COLOR’ was not declared in this scope
293 | cv_img = cv::imread(image.first, CV_LOAD_IMAGE_COLOR);
|