ubuntu20 caffe opencv4 conda python3.7.10
如果你只想要pycaffe，那可以conda+python3.7.10，然后conda install caffe-gpu
环境
ubuntu20
cuda11.7 cudnn8（不过caffe似乎不能加速cudnn8）
conda python3.7.10
假设你已经装了opencv4，cuda11.7，cudnn8
opencv4安装教程
cuda和cudnn安装教程
建议ubuntu和anaconda，pip都换个源（ubuntu换源，anaconda换源，建议ustc/bfsu）
sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler -y
sudo apt-get install --no-install-recommends libboost-all-dev -y
sudo apt-get install libopenblas-dev liblapack-dev libatlas-base-dev -y
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev -y
sudo apt-get install git cmake build-essential -y
python
conda create -n caffe_env python==3.7.10 pip setuptools
conda activate caffe_env
pip install numpy scipy scikit-image matplotlib ipython h5py leveldb networkx nose pandas python-dateutil protobuf python-gflags pyyaml pillow six google-api-python-client
conda install pydot graphviz boost
这里用conda装boost是因为apt似乎只会装和系统python一样的版本的boost
安装
https://github.com/BVLC/caffe
tar -zxvf caffe-master.zip
cd caffe-master
cp Makefile.config.example Makefile.config
修改Makefile.config
修改代码：
#USE_OPENCV := 0
#OPENCV_VERSION := 3
#WITH_PYTHON_LAYER := 1
为：
USE_OPENCV := 1
OPENCV_VERSION := 4
WITH_PYTHON_LAYER := 1
修改代码：
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib
为：
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial /usr/local/include/opencv4
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial
# 最后是自己的opencv的include和bin路径，解决fatal error: opencv2/core/core.hpp问题
修改代码：
# ANACONDA_HOME := $(HOME)/anaconda
# PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
# $(ANACONDA_HOME)/include/python2.7 \
# $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include
为：
ANACONDA_HOME := $(HOME)/miniconda3/envs/caffe_env
PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
$(ANACONDA_HOME)/include/python3.7m \
$(ANACONDA_HOME)/lib/python3.7/site-packages/numpy/core/include
# 根据自己的路径修改
修改代码：
# PYTHON_LIBRARIES := boost_python3 python3.5m
修改代码：
# ANACONDA_HOME := $(HOME)/anaconda
# PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
# $(ANACONDA_HOME)/include/python2.7 \
# $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include
为：
PYTHON_LIBRARIES := boost_python37 python3.7m
# 根据自己的路径修改
修改代码：
CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
-gencode arch=compute_20,code=sm_21 \
-gencode arch=compute_30,code=sm_30 \
-gencode arch=compute_35,code=sm_35 \
-gencode arch=compute_50,code=sm_50 \
-gencode arch=compute_52,code=sm_52 \
-gencode arch=compute_60,code=sm_60 \
-gencode arch=compute_61,code=sm_61 \
-gencode arch=compute_61,code=compute_61
为：
CUDA_ARCH := -gencode arch=compute_50,code=sm_50 \
-gencode arch=compute_52,code=sm_52 \
-gencode arch=compute_60,code=sm_60 \
-gencode arch=compute_61,code=sm_61 \
-gencode arch=compute_61,code=compute_61
# 根据自己的架构和版本修改，可参考https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
修改Makefile
修改Makefile
修改代码：
NVCCFLAGS +=-ccbin=$(CXX) -Xcompiler-fPIC $(COMMON_FLAGS)
为：
NVCCFLAGS += -D_FORCE_INLINES -ccbin=$(CXX) -Xcompiler -fPIC $(COMMON_FLAGS)
修改代码：
LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_hl hdf5
为：
LIBRARIES += glog gflags protobuf boost_system boost_filesystem hdf5_hl hdf5 m opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs
# 解决undefined reference to cv::imread(cv::String const&, int)...,可参考https://github.com/BVLC/caffe/issues/2348
修改代码：
PYTHON_LIBRARIES ?= boost_python python2.7
为：
PYTHON_LIBRARIES ?= boost_python37 python3.7
# 根据自己的版本修改
修改src/caffe/util/io.cpp
注意这里要修改的可能不止一处
将：
CV_LOAD_IMAGE_COLOR
修改为：
cv::IMREAD_COLOR
将：
CV_LOAD_IMAGE_GRAYSCALE
修改为：
cv::IMREAD_GRAYSCALE
修改src/caffe/layers/window_data_layer.cpp
注意这里要修改的可能不止一处
将：
CV_LOAD_IMAGE_COLOR
修改为：
cv::IMREAD_COLOR
将：
CV_LOAD_IMAGE_GRAYSCALE
修改为：
cv::IMREAD_GRAYSCALE
修改src/caffe/data_transformer.cpp
修改代码：
#ifdef USE_OPENCV
#include include/caffe/common.hpp
新增代码：
// Supporting OpenCV4
#if (CV_MAJOR_VERSION == 4)
#define CV_LOAD_IMAGE_COLOR cv::IMREAD_COLOR
#define CV_LOAD_IMAGE_GRAYSCALE cv::IMREAD_GRAYSCALE
#endif
以下为原代码（用于定位）：
// See PR #1236
namespace cv { class Mat; }
编译
先切一下环境（也许不需要）
conda activate caffe_env
make all -j8
测试
先切一下环境（也许不需要）
conda activate caffe_env
这里是因为boost_python是conda装的，默认不在系统里，所以先改一下LD_LIBRARY_PATH
export LD_LIBRARY_PATH="/home/icml/miniconda3/envs/caffe_env/lib:$LD_LIBRARY_PATH"
make runtest -j8
编译pycaffe
先切一下环境（也许不需要）
conda activate caffe_env
需要的python包在最前面创建python的时候就已经安装了
make pycaffe -j8
测试
先切一下环境
conda activate caffe_env
需要的python包在最前面创建python的时候就已经安装了
这里是因为boost_python是conda装的，默认不在系统里，所以先改一下LD_LIBRARY_PATH
export LD_LIBRARY_PATH="/home/icml/miniconda3/envs/caffe_env/lib:$LD_LIBRARY_PATH"
然后
export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION='python'
这一句是为了防止出现下面这个错误
开始测试
make pytest -j8
修改环境变量
在~/.bashrc的后面加入
export LD_LIBRARY_PATH="/home/icml/miniconda3/envs/caffe_env/lib:$LD_LIBRARY_PA>
export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION='python'
export PYTHONPATH="${PYTHONPATH}:/mnt/data/caffe-master/python"
然后
source ~/.bashrc
如果你有zsh，就再做一遍类似的操作
测试
C++
mnist
首先cd进入caffe的目录
sh data/mnist/get_mnist.sh
sh examples/mnist/create_mnist.sh
nano examples/mnist/lenet_solver.prototxt
修改最后的max_iter和solver_mode（虽然我没改
开始训练
time sh examples/mnist/train_lenet.sh
花了1min，还行把
https://blog.csdn.net/ynzzxc/article/details/116780168
https://blog.csdn.net/yhaolpz/article/details/71375762
https://qengineering.eu/install-caffe-on-ubuntu-20.04-with-opencv-4.4.html
https://stackoverflow.com/questions/50839667/protofile-proto-a-file-with-this-name-is-already-in-the-pool