ubuntu16.04下安裝caffe

最近在學機器視覺和深度學習，需要安裝caffe。苦於之前用的fedora安裝依賴各種安裝不上，索性下載了最新版ubuntu16.04來安裝caffe。不過由於系統比較新，有些依賴庫名字已經變了，安裝的時候需要修改Makefile和配置文件。爲了讓其他人少走彎路，總結了下自己的經驗。

1.下載caffe源碼

git clone htts://github.com/BVLC/caffe.git

2.安裝依賴庫

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libboost-all-dev libhdf5-serial-dev libgflags-dev libgoogle-glog-dev liblmdb-dev protobuf-compiler libopenblas-dev

3.修改Makefile

cd caffe
cp Makefile.config.example Makefile.config

修改Makefile.config如下幾個配置：

USE_CPU := 1
USE_OPENCV := 1
USE_LMDB := 1
BLAS := open
WITH_PYTHON_LAYER := 1

修改Makefile如下幾個地方：

diff --git a/Makefile b/Makefile
index 2489406..5d56738 100644
--- a/Makefile
+++ b/Makefile
@@ -171,14 +171,14 @@ ifneq ("$(wildcard $(CUDA_DIR)/lib64)","")
 endif
 CUDA_LIB_DIR += $(CUDA_DIR)/lib

-INCLUDE_DIRS += $(BUILD_INCLUDE_DIR) ./src ./include
+INCLUDE_DIRS += $(BUILD_INCLUDE_DIR) ./src ./include /usr/include/hdf5/serial
 ifneq ($(CPU_ONLY), 1)
        INCLUDE_DIRS += $(CUDA_INCLUDE_DIR)
        LIBRARY_DIRS += $(CUDA_LIB_DIR)
        LIBRARIES := cudart cublas curand
 endif

-LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_hl hdf5
+LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_serial_hl hdf5_serial

 # handle IO dependencies
 USE_LEVELDB ?= 1

將減號（-）那行用加號（+）那行替換掉，不要帶第一個加號

4.編譯

make all
make test
make runtest

5.編譯python庫

sudo apt-get install python-pip python-dev build-essential
sudo pip install --upgrade pip
sudo pip install -r python/requirements.txt
make pycaffe

6.安裝並運行notebook

sudo apt-get install ipython-notebook python-sympy
sudo pip install jupyter
mkdir notebook
cd notebook
ipython notebook

7.運行caffe

cd caffe
./data/mnist/get_mnist.sh
./example/mnist/create_mnist.sh
./example/mnist/train_lenet.sh

不過執行 train_lenet.sh 的時候始終過不去，貌似是 tools/caffe.cpp 的問題，沒有正確處理CPU模式和GPU模式
下面這篇文章裏有比較詳細的介紹
http://www.open-open.com/lib/view/open1421995285109.html