【无标题】

• 原文下载： https://arxiv.org/pdf/1506.06579.pdf
• 论文翻译：可解释论文阅读（一）---Understanding Neural Networks Through Deep Visualization_simonyan et al. 2013_pure water的博客-CSDN博客 
• 论文翻译：Understanding Neural Networks Through Deep Visualization 论文笔记_Jayxbx的博客-CSDN博客
• 工具配置教程：[Deconv]|[cnn]Deep visualization toolbox安装_御前上等的博客-CSDN博客

---

0. Ubuntu 20.04 安装并编译opencv4.7.0

1. opencv下载地址：https://opencv.org/releases.html
2. 按照如下的步骤编译； 

 

问题①：找不到opencv包
解决办法：Package opencv was not found in the pkg-config search path._stdcoutzrh的博客-CSDN博客
        ①在/usr/local/lib路径下新建pkgconfig文件夹，编辑opencv.pc文件，里面写上opencv的路径
        ②将Opencv的路径写入bashrc，export  PKG_CONFIG_PATH=/usr/local/lib/pkgconfig
        ③source ~/.bashrc 激活

还是找不到opencv包，于是Makefile.config文件中的include_dirs中，加入opencv的路径
成功解决！

# Whatever else you find you need goes here.
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial /usr/local/include/opencv4
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib

问题②：PROTOC src/caffe/proto/caffe.proto make：protoc：命令未找到 make: ***  [Makefile:630：.build_release/src/caffe/proto/caffe.pb.cc] 错误 127

caffe编译错误：make: protoc: Command not found_陈 洪 伟的博客-CSDN博客

解决办法：sudo apt-get install  protobuf-compiler

问题③：未安装boost
解决办法：在Ubuntu上安装Boost的五种方法(全网最全，建议收藏)_ubuntu安装boost_小熊coder的博客-CSDN博客

sudo apt-get update
sudo apt-get install libboost-all-dev

问题④：opencv4的问题，改源码
src/caffe/layers/window_data_layer.cpp:293:42: error: ‘CV_LOAD_IMAGE_COLOR’ was not declared in this_陈 洪 伟的博客-CSDN博客

解决办法：cv::IMREAD_COLOR替代CV_LOAD_IMAGE_COLOR   cv::IMREAD_GRAYSCALE

问题⑤：编译caffe遇到的问题（二）fatal error: leveldb/db.h: No such file or directory_muyeluo123的博客-CSDN博客

 解决办法：先把子模块都拉取下来 git submodule / init / update; 然后切换到build文件夹，进行cmake

 Git submodule 仓库中包含另外的仓库(子仓库)_weixin_30918633的博客-CSDN博客

        ①git submodule 有减号
        ②git submodule init 
        ③git submodule 无减号
        ④git submodule update 拉取子仓库下来
        ⑤mkdir -p build && cd build
        ⑥cmake -DCMAKE_BUILD_TYPE=Release .. && cmake --build .
        ⑦make
        ⑧将include下的头文件放到/usr/local/include/中去

问题⑤：make：/usr/local/cuda/bin/nvcc：命令未找到
解决办法：错误make: nvcc: Command not found 解决方案-CSDN博客

 1. cuda linux下载地址

CUDA Toolkit 12.1 Downloads | NVIDIA Developer

 算力匹配表：CUDA计算能力&显卡对照表_cuda算力_Janus_V的博客-CSDN博客

nvcc fatal : Unsupported gpu architecture 'compute_20'_青春如诗_的博客-CSDN博客

hdf5的路径，要将/usr/local/hdf5/include给放到Makefile.config的include路径中去。 

linux - /usr/bin/ld: 找不到 -lpthreads - SegmentFault 思否s​​​​​​​s

snappy安装

/usr/bin/ld: 找不到 -lhdf5_hl